Hybrid path planning for mobile robots in complex environments: Fusing improved BI-RRT and enhanced DWA Authors: Guogang Wang, Hongwei Sun, Zichao Feng Published: 2026 ABSTRACT Mobile robot navigation in complex environments requires efficient global planning and reactive local obstacle avoidance. This paper proposes a hybrid path planning algorithm that integrates an improved Bidirectional Rapidly-exploring Random Tree (BI-RRT) for global guidance with an enhanced Dynamic Window Approach (DWA) for local execution. The BI-RRT algorithm incorporates a multi-sampling point strategy, dynamic step-size adjustment, and an Artificial Potential Field (APF)-inspired node selection mechanism to reduce sampling redundancy and improve path quality. The DWA evaluation function is augmented with historical trajectory information and dynamic obstacle prediction to enable smoother avoidance of moving obstacles. The hierarchical architecture establishes a continuous transition from global path optimization to local obstacle avoidance, with the global BI-RRT path providing waypoints for the local DWA controller. Experimental validation was conducted through MATLAB simulations across three environmental scales (50×50, 100×100, and 200×200) with obstacle densities ranging from 15% to 25%, ROS-based tests, and physical experiments on a Jetauto robot platform. Compared to baseline methods including RRT*, BI-RRT, APF-RRT*, and Informed-RRT*, the proposed algorithm improves search efficiency, achieving an 88–93% reduction in sampling nodes and 89–97% reduction in computation time compared to standard RRT*, while maintaining 100% success rate across tested environments. Physical robot experiments demonstrate 8.17% reduction in path length, 13.85% reduction in execution time, and 40% reduction in turning maneuvers compared to conventional BI-RRT. The algorithm maintains highly reliable navigation performance across the specifically evaluated test environments, including dynamic obstacle scenarios, demonstrating its viability under the tested conditions. 1. Introduction Mobile robots are widely used in industrial applications due to their autonomous navigation and real-time decision-making capabilities, which improve operational efficiency. Path planning, a core research area in mobile robotics [1,2], is generally categorized into global and local planning based on the robot’s environmental awareness. Global path planning assumes complete map knowledge and computes an optimal path in a static environment. Commonly used algorithms include Dijkstra [3], A* [4], ant colony optimization [5,6], and rapidly-exploring random trees (RRT) [7,8]. As a sampling-based method, RRT efficiently explores feasible paths by constructing a random tree, adapting well to complex environments. However, conventional RRT-based algorithms suffer from high computational cost, suboptimal paths, and slow convergence. Recent improvements include bidirectional RRT* (BI-RRT*) for robotic arm planning [9] and heuristic-integrated Bi-RRT with Dijkstra optimization [10]. While these methods enhance efficiency and path quality, challenges remain in dynamic or densely cluttered settings. Local path planning methods, such as artificial potential field (APF) [11], time elastic band (TEB) [12], and dynamic window approach (DWA) [13], handle real-time obstacle avoidance. DWA offers flexibility, low computational overhead, and ease of implementation. By comprehensively accounting for the robot’s kinematic constraints and the environmental map, the DWA constructs a feasible velocity window to generate multiple candidate trajectories. Subsequently, it identifies the optimal trajectory and the corresponding velocity commands through a robust evaluation function. Nevertheless, in complex dynamic or unstructured environments, DWA, like other local planners, is prone to local minima, leading to planning failure. To overcome these limitations, this paper proposes a hybrid algorithm that combines an improved BI-RRT with an optimized DWA. The key contribution of this work lies in the systematic coupling mechanism within the proposed fusion framework, rather than the simple superposition of isolated algorithms.The specific innovations and contributions can be summarized in the following three points: Coupled Mechanism in the Global Planning Layer: Rather than operating independently, the multi-sampling scoring strategy, artificial potential field (APF) guidance, and dynamic step-size adjustment share unified local obstacle density computations. Together, they form a closed-loop operation within each iteration that systematically integrates competitive sampling evaluation, adaptive step-size calculation, and expansion direction constraint. Coupled Mechanism in the Global Planning Layer: Rather than operating independently, the multi-sampling scoring strategy, artificial potential field (APF) guidance, and dynamic step-size adjustment share unified local obstacle density computations. Together, they form a closed-loop operation within each iteration that systematically integrates competitive sampling evaluation, adaptive step-size calculation, and expansion direction constraint. Hierarchical Synergy and Enhanced Local Control: The enhanced dynamic window approach (DWA) embeds both historical trajectory consistency and dynamic obstacle prediction costs into a unified evaluation function, utilizing a strict weight hierarchy to balance trajectory smoothness and navigation safety. Moreover, the proposed architecture establishes a hierarchical synergy, facilitating a continuous transition from the topologically optimized global guidance of the ADBI-RRT to the high-frequency, dynamic obstacle avoidance of DWA. Hierarchical Synergy and Enhanced Local Control: The enhanced dynamic window approach (DWA) embeds both historical trajectory consistency and dynamic obstacle prediction costs into a unified evaluation function, utilizing a strict weight hierarchy to balance trajectory smoothness and navigation safety. Moreover, the proposed architecture establishes a hierarchical synergy, facilitating a continuous transition from the topologically optimized global guidance of the ADBI-RRT to the high-frequency, dynamic obstacle avoidance of DWA. Integrated Performance Validation: The proposed algorithm achieves substantial performance improvements over baseline methods, including an 88–93% reduction in sampling nodes and an 89–97% decrease in computation time across varied test scenarios. These gains result from the collective operation of the coupled modules within the fusion framework. However, the current experimental design validates the complete algorithm as a whole rather than decomposing the individual contributions of each module through controlled ablation studies. Therefore, while the three improvements (multi-sampling, APF guidance, dynamic stepping) are designed to work synergistically, the specific contribution weight of each component to the overall performance enhancement has not been quantitatively isolated. Integrated Performance Validation: The proposed algorithm achieves substantial performance improvements over baseline methods, including an 88–93% reduction in sampling nodes and an 89–97% decrease in computation time across varied test scenarios. These gains result from the collective operation of the coupled modules within the fusion framework. However, the current experimental design validates the complete algorithm as a whole rather than decomposing the individual contributions of each module through controlled ablation studies. Therefore, while the three improvements (multi-sampling, APF guidance, dynamic stepping) are designed to work synergistically, the specific contribution weight of each component to the overall performance enhancement has not been quantitatively isolated. The remainder of this paper is organized as follows: Section 2 describes the ADBI-RRT algorithm. Section 3 presents the enhanced DWA. Section 4 provides simulation and experimental results. Section 5 concludes the paper and suggests future work. 2. Improved BI-RRT algorithm 2.1 Traditional BI-RRT and its limitations The Bidirectional Rapidly-exploring Random Tree (BI-RRT) algorithm [14] accelerates spatial exploration by growing two search trees simultaneously from the start and goal states until they connect. Traditional BI-RRT suffers from three limitations: blind uniform sampling generates numerous invalid nodes in obstacle-dense regions, fixed step size causes collisions in narrow corridors and slow progress in open areas, and lack of heuristic guidance produces jagged suboptimal trajectories. The ADBI-RRT framework addresses these issues through adaptive sampling, dynamic step-size adjustment, and APF-guided node selection. 2.2 Proposed improvements 2.2.1 Multi-random sampling strategy. The traditional BI-RRT single-sampling strategy suffers from slow convergence and excessive randomness. The proposed multi-sampling approach generates multiple candidate points within the planning space and selects the optimal one to guide tree expansion, improving both sampling efficiency and goal orientation. nherent limitations of the single-sampling To quantitatively evaluate the relative advantages and disadvantages of each candidate sampling point, a dedicated scoring function is designed, which incorporates three key influential factors: obstacle density around the sampling point, the shortest distance from the sampling point to the nearest obstacle, and the vertical distance from the sampling point to the start-goal line (i.e., the line connecting the start and goal configurations). The optimal sampling point is determined by calculating and comparing the score of each candidate. For N randomly generated candidate sampling points, the score of the i -th point score_i is computed using Eq. (1): where scorei denotes the sampling score of the i-th candidate point, ρi represents the obstacle density at the i -th sampling point, min_distance is the shortest distance from the i -th sampling point to the nearest obstacle, area_size denotes the size of the region used for obstacle density calculation, point_line_distancei is the vertical distance from the i -th sampling point to the line connecting the start point and the target point, line_length represents the length of the line connecting the start point and the target point, and c is a weighting coefficient that adjusts the relative importance of the distance term. 2.2.2 Obstacle density calculation and implementation. The obstacle density ρi quantifies the local obstacle concentration within a neighborhood region centered at sampling point pi. This metric directly influences both the sampling probability and step size adaptation. The calculation procedure is defined as follows. 2.2.2.1 Mathematical definition: For a sampling point pi at coordinates (xi,yi), the obstacle density is computed by counting obstacle-occupied cells within a square region of side length area_size: where Nobs(pi) is the number of obstacle-occupied grid cells within the region ℛ(pi)={(x,y):|x−xi|≤area_size/2,|y−yi|≤area_size/2}, and Ntotal(pi)=(area_size/Δgrid)2 is the total number of cells in the region, with Δgrid being the grid resolution. 2.2.2.2 Implementation steps: The density calculation follows this procedure: Define square region ℛ centered at pi with side length area_size=5.0. Define square region ℛ centered at pi with side length area_size=5.0. Discretize the map into a uniform grid with resolution Δgrid=0.1. Discretize the map into a uniform grid with resolution Δgrid=0.1. Count obstacle-occupied cells: Nobs=∑(x,y)∈ℛ𝕀[map(x,y)=1], where 𝕀[·] is the indicator function. Count obstacle-occupied cells: Nobs=∑(x,y)∈ℛ𝕀[map(x,y)=1], where 𝕀[·] is the indicator function. Compute normalized density: ρi=Nobs/Ntotal∈[0,1]. Compute normalized density: ρi=Nobs/Ntotal∈[0,1]. 2.2.2.3 Computational example: Consider pi=(10.0,15.0) in a 50×50 environment with area_size=5.0 and Δgrid=0.1. The region ℛ spans (7.5, 12.5) to (12.5, 17.5), containing Ntotal=(5.0/0.1)2=2500 cells. If Nobs=375 cells are occupied, then ρi=375/2500=0.15 (15% obstacle coverage). 2.2.2.4 Parameter selection: The value area_size=5.0 was determined through systematic parameter sweeps. Smaller values (e.g., 2.0) yield overly localized estimates, while larger values (e.g., 10.0) produce over-smoothed density fields. The selected value balances local sensitivity and computational efficiency across all test environments (50×50, 100×100, 200×200). Table 1 summarizes all parameters in Eq. (1). 10.1371/journal.pone.0357770.t001Table 1Variable definitions for multi-sampling scoring function (Eq. 1).SymbolDescriptionValue/UnitPhysical Meaning scorei Sampling score–Quantifies candidate point quality; higher is better ρi Obstacle density[0, 1]Fraction of obstacle-occupied cells in local region min_distance Distance to nearest obstacleMinimum clearance from pi to any obstacle area_size Density calculation region5.0Side length of square search region Δgrid Grid resolution0.1Discretization resolution for occupancy map min_dist Minimum safe distance0.5Safety threshold: robot radius (0.3) + margin (0.2) Variable definitions for multi-sampling scoring function (Eq. 1). 2.2.2.5 Scoring function rationale: Eq. (1) penalizes candidates in high-density regions (first term) and far from the start-goal line (second term). The exponential decay ensures smooth score variation. Points with min_distance20.0): Excessive repulsion causes the robot to maintain unnecessarily large clearances from obstacles, increasing path length by 20–30% and reducing navigation efficiency. Upper bound (kr>20.0): Excessive repulsion causes the robot to maintain unnecessarily large clearances from obstacles, increasing path length by 20–30% and reducing navigation efficiency. Optimal range (kr∈[12.0,18.0]): Provides robust collision avoidance with minimal path deviation. The value kr=15.0 achieved zero collision failures across all test scenarios while maintaining near-optimal path lengths. Optimal range (kr∈[12.0,18.0]): Provides robust collision avoidance with minimal path deviation. The value kr=15.0 achieved zero collision failures across all test scenarios while maintaining near-optimal path lengths. Influence Distance (ρ0=2.5 grid units): The influence threshold defines the spatial extent of the repulsive field in the normalized workspace. This value was selected to provide sufficient advance warning for obstacle avoidance. When deployed in physical environments with scale factor s = 1.0 m/unit, this corresponds to 2.5 m physical distance; the dimensionless parameter value remains constant across different deployment scales. This value was selected based on the robot’s maximum velocity (1.0 m/s) and reaction time (2.0 s), ensuring sufficient advance warning to decelerate and avoid obstacles. Smaller values (ρ0<2.0 m) resulted in late obstacle detection, while larger values (ρ0>3.0 m) caused premature path adjustments and increased computational load. Distance Exponent (n = 2): The quadratic distance term ρgn(p,pgoal) in Eq. (12) ensures that repulsive forces are stronger when the robot is far from the goal, preventing the robot from being pushed away from the target region. This design addresses the goal non-reachable problem (GNRP) commonly observed in traditional APF methods. Force Decomposition Rationale: The repulsive force is decomposed into obstacle-directed (𝐅r1) and goal-directed (𝐅r2) components to simultaneously achieve obstacle avoidance and goal convergence. The relative magnitudes of these components are automatically adjusted based on the robot’s distance to both the obstacle and the goal, ensuring adaptive behavior in different scenarios. Table 5 summarizes the parameter values for the repulsive potential field. 10.1371/journal.pone.0357770.t005Table 5Parameter values for repulsive potential field (Eqs. 12–14).ParameterDescriptionValueJustification kr Repulsive gain coefficient15.0 (dimensionless)Ensures zero collision failures; tuned via safety analysis ρ0 Influence distance threshold2.5 grid unitsDefines local repulsive field extent; compatible with normalized workspace n Distance exponent2Quadratic term prevents GNRP; standard APF formulation Parameter values for repulsive potential field (Eqs. 12–14). 3. Enhanced DWA algorithm 3.1 Traditional evaluation function The DWA algorithm is a local path planning algorithm based on velocity sampling. Within the velocity sampling space, multiple candidate trajectories are generated. G(v,ω) is the evaluation function of the original DWA algorithm. The optimal trajectory is obtained by evaluating and scoring each candidate trajectory. The trajectory evaluation function expression under the velocity (v,ω) is G(v,ω) is the evaluation function of the original DWA algorithm. the optimal trajectory is obtained by evaluating the function for scoring, and the trajectory evaluation function expression under the velocity (v,ω) is Where α, β, γ are the weighting coefficients of the evaluation function, Heading(v,ω) is used to calculate the robot azimuth angle, Obst(v,ω) is used to calculate the distance between the end trajectory and the obstacle, and velocity(v,ω) is used to calculate the current velocity magnitude. 3.1.1 Parameter selection and justification for DWA weights. The weighting coefficients in the DWA evaluation function were systematically tuned to balance multiple competing objectives: goal-directedness, obstacle avoidance, and velocity optimization. Heading Weight (α=0.5): This coefficient controls the importance of aligning the robot’s trajectory with the goal direction. The value α=0.5 was selected to provide moderate goal-directedness without over-constraining the robot’s ability to deviate for obstacle avoidance. Lower values (α<0.3) resulted in meandering paths with 15–20% increased path length, while higher values (α>0.7) reduced obstacle avoidance effectiveness, increasing collision risk by 5–8%. Obstacle Weight (Baseline Mode, β=0.8): This coefficient prioritizes collision avoidance by penalizing trajectories that pass close to obstacles. The relatively high value β=0.8 (compared to α and γ) reflects the critical importance of safety in robotic navigation. This value was validated through safety analysis, achieving zero collision failures in static environments and less than 2% collision rate in highly dynamic scenarios. Velocity Weight (γ=0.3): This coefficient encourages the robot to maintain higher velocities when safe, improving navigation efficiency. The value γ=0.3 was selected to promote speed without compromising safety. Higher values (γ>0.5) caused the robot to maintain excessive speeds near obstacles, while lower values (γ<0.2) resulted in overly conservative behavior with 25–30% increased travel time. Weight Normalization: The weights are normalized such that α+β+γ=1.6, ensuring consistent scaling across different scenarios. This normalization facilitates parameter transfer to different robot platforms and environments. Table 6 summarizes the basic DWA parameter configuration. 10.1371/journal.pone.0357770.t006Table 6Parameter configuration for basic DWA evaluation function (Eq. 14).SymbolDescriptionValueJustification α Heading weight0.5Balances goal-directedness and flexibility β Obstacle weight (Baseline Mode)0.8Prioritizes safety in static environments γ Velocity weight0.3Encourages efficiency without compromising safety Heading(v,ω) Heading costNormalized [0,1]Angular deviation from goal direction Obst(v,ω) Obstacle costNormalized [0,1]Inverse of minimum clearance distance Velocity(v,ω) Velocity costNormalized [0,1]Ratio of current to maximum velocity Parameter configuration for basic DWA evaluation function (Eq. 14). 3.2 Proposed enhancements 3.2.1 Historical trajectory information fusion. In this paper, the Historical_tra(v,ω) cost consists of two components: direction consistency cost and curvature cost, denoted as Historical_dir and Historical_curve respectively. The direction consistency cost is adopted to compute the discrepancy between the historical trajectory direction and the predicted trajectory direction, as presented in Equation (15). Correspondingly, the curvature cost, defined in Equation (16), quantifies the smoothness of the predicted trajectory by evaluating the cumulative angular changes between consecutive trajectory points, which is crucial for avoiding abrupt turns and ensuring motion stability within the DWA framework. The overall historical trajectory cost, formulated in Equation (17), integrates these two components with weighted coefficients to comprehensively evaluate the consistency and feasibility of the predicted trajectory. where θhdenotes the historical trajectory direction and θr represents the predicted trajectory direction. The curvature cost is introduced to quantify the degree of curvature of the predicted trajectory, thereby mitigating the occurrence of sharp turns, as mathematically expressed in Equation (17).m denotes the total number of trajectory points, i.e., the trajectory is composed of m discrete points.The orientation θi at each trajectory point is utilized to characterize the composition of the overall Historical_tra(v,ω) cost, which is formulated as a weighted linear combination of the direction cost and the curvature cost (Equation (18)), with ρ1 and ρ1 serving as the corresponding weighting coefficients. 3.2.1.1 Parameter selection and justification for historical trajectory weights: The historical trajectory weights were tuned to ensure smooth motion while maintaining responsiveness to environmental changes. Direction Consistency Weight (ρ1=0.6): This coefficient penalizes abrupt changes in trajectory direction, promoting smooth motion. The value ρ1=0.6 was selected to maintain directional stability without causing excessive rigidity. Lower values (ρ1<0.4) resulted in jerky motion with frequent direction changes, reducing passenger comfort and increasing actuator wear. Higher values (ρ1>0.8) caused the robot to be overly committed to historical directions, reducing responsiveness to newly detected obstacles. Curvature Weight (ρ2=0.4): This coefficient penalizes high-curvature trajectories, encouraging smooth paths with gentle turns. The value ρ2=0.4 was selected to avoid sharp turns while allowing necessary maneuvers in constrained spaces. The ratio ρ1:ρ2=0.6:0.4 reflects the relative importance of maintaining directional consistency over minimizing curvature. Adaptive Weight Adjustment: In highly dynamic environments with frequent obstacle movements, the historical weights are automatically reduced by 30–50% to prioritize real-time obstacle avoidance over trajectory smoothness. This adaptive mechanism is triggered when the rate of environmental change exceeds a predefined threshold (3 new obstacles detected per second). Trajectory Memory Length: The historical trajectory buffer maintains the most recent 50 trajectory points (corresponding to approximately 2.5 seconds of motion at 20 Hz control frequency). This window length balances short-term smoothness and long-term adaptability. Table 7 summarizes the historical trajectory parameter configuration. 10.1371/journal.pone.0357770.t007Table 7Parameter configuration for historical trajectory evaluation (Eqs. 15–17).SymbolDescriptionValueJustification ρ1 Direction consistency weight0.6Promotes smooth directional changes ρ2 Curvature weight0.4Penalizes sharp turns; secondary to direction θh Historical trajectory directionVariable (rad)Computed from previous valid trajectory θr Predicted trajectory directionVariable (rad)Computed from current velocity sample m Number of trajectory points30Prediction horizon: 3.0 s at 10 Hz sampling θi Orientation at point iVariable (rad)Discrete heading angles along trajectoryBuffer sizeHistorical trajectory memory50 pointsCorresponds to 2.5 s at 20 Hz control rate Parameter configuration for historical trajectory evaluation (Eqs. 15–17). 3.2.2 Dynamic obstacle prediction. A dynamic obstacle prediction mechanism is introduced to address DWA limitations in dynamic environments. Given current position (x0,y0), velocity (vx,vy), and acceleration (ax,ay), the predicted position at time t is: For each point Di on the predicted trajectory within the time horizon T, the minimum distance to the predicted obstacle position is calculated as Based on the above minimum distances, the dynamic obstacle cost function Dynamicdist(𝐯,ω) is defined as where λ denotes a tiny positive constant (empirically set to 10−6) to avoid division-by-zero errors in numerical computation. Distinct from the conventional DWA evaluation function, the proposed improved version incorporates two additional critical components: historical trajectory consistency and dynamic obstacle prediction. Accordingly, the enhanced evaluation function can comprehensively balance temporal trajectory continuity and dynamic environmental constraints. The final improved evaluation function G* is defined as follows: Where τ represents the weight coefficient for historical trajectory fusion, which is set to 0.8 in this study; notably, this weight decreases as the complexity of the robot’s operating environment increases to prioritize real-time obstacle avoidance over trajectory continuity. μ denotes the weight coefficient for dynamic obstacle cost, which is set to 2.3 to emphasize the priority of dynamic obstacle avoidance in cluttered scenarios. 3.2.2.1 Emergency stop mechanism for unpredictable obstacle behavior: While the dynamic obstacle prediction model in Eq. (19) assumes constant acceleration motion, real-world obstacles may exhibit unpredictable behaviors such as sudden direction changes, abrupt stops, or erratic movements that deviate from the predicted trajectory. To ensure safety in such scenarios, an emergency stop mechanism is integrated into the control framework. Prediction Error Monitoring: At each control cycle, the actual observed position of the dynamic obstacle (xobs,yobs) is compared with its predicted position (xt,yt) from Eq. (19). The prediction error Epred is computed as: Emergency Stop Trigger Condition: If the prediction error exceeds a predefined safety threshold Ethresh (empirically set to 0.5 m based on the robot’s sensing accuracy and reaction time), the obstacle is classified as exhibiting unpredictable behavior. The emergency stop condition is defined as: Emergency Response Protocol: When the emergency stop condition is triggered: Immediate Deceleration: The robot’s linear and angular velocities are set to zero within one control cycle (Δt=50 ms), ensuring a rapid halt. Immediate Deceleration: The robot’s linear and angular velocities are set to zero within one control cycle (Δt=50 ms), ensuring a rapid halt. Trajectory Re-evaluation: The DWA local planner suspends normal operation and enters a monitoring state, continuously tracking the unpredictable obstacle’s motion. Trajectory Re-evaluation: The DWA local planner suspends normal operation and enters a monitoring state, continuously tracking the unpredictable obstacle’s motion. Safe Distance Verification: The robot remains stationary until the minimum distance dmin to the unpredictable obstacle exceeds a safe threshold dsafe (set to 1.5 m, accounting for the robot’s footprint and a safety margin). Safe Distance Verification: The robot remains stationary until the minimum distance dmin to the unpredictable obstacle exceeds a safe threshold dsafe (set to 1.5 m, accounting for the robot’s footprint and a safety margin). Motion Resumption: Once dmin>dsafe and the obstacle’s motion becomes predictable again (i.e., Epreddsafe and the obstacle’s motion becomes predictable again (i.e., Epred1.0): Excessive historical influence causes the robot to be overly committed to previous trajectories, reducing obstacle avoidance effectiveness and increasing collision risk by 6–9% in dynamic environments. Upper bound (τ>1.0): Excessive historical influence causes the robot to be overly committed to previous trajectories, reducing obstacle avoidance effectiveness and increasing collision risk by 6–9% in dynamic environments. Optimal range (τ∈[0.7,0.9]): Balances smoothness and responsiveness. The value τ=0.8 achieved the best trade-off, reducing trajectory jitter by 65% while maintaining collision-free navigation. Optimal range (τ∈[0.7,0.9]): Balances smoothness and responsiveness. The value τ=0.8 achieved the best trade-off, reducing trajectory jitter by 65% while maintaining collision-free navigation. Adaptive Adjustment: In highly dynamic environments (more than 3 moving obstacles detected), τ is automatically reduced to 0.5 to prioritize real-time obstacle avoidance over trajectory smoothness. This adaptive mechanism is critical for maintaining safety in unpredictable scenarios. Dynamic Obstacle Weight (μ=2.3): This coefficient emphasizes the importance of dynamic obstacle avoidance, which is more critical than static obstacle avoidance due to the unpredictability of moving obstacles. Tuning Process: Lower bound (μ<1.5): Insufficient emphasis on dynamic obstacles leads to collision failures, particularly when obstacles move at velocities comparable to or exceeding the robot’s velocity (0.5–1.0 m/s). Collision rate increased to 8–12% in preliminary tests. Lower bound (μ<1.5): Insufficient emphasis on dynamic obstacles leads to collision failures, particularly when obstacles move at velocities comparable to or exceeding the robot’s velocity (0.5–1.0 m/s). Collision rate increased to 8–12% in preliminary tests. Upper bound (μ>3.0): Excessive emphasis causes overly conservative behavior, with the robot frequently stopping or taking unnecessarily long detours, increasing travel time by 35–45%. Upper bound (μ>3.0): Excessive emphasis causes overly conservative behavior, with the robot frequently stopping or taking unnecessarily long detours, increasing travel time by 35–45%. Optimal range (μ∈[2.0,2.5]): Provides robust dynamic obstacle avoidance while maintaining navigation efficiency. The value μ=2.3 achieved less than 2% collision rate in highly dynamic scenarios while maintaining reasonable travel times. Optimal range (μ∈[2.0,2.5]): Provides robust dynamic obstacle avoidance while maintaining navigation efficiency. The value μ=2.3 achieved less than 2% collision rate in highly dynamic scenarios while maintaining reasonable travel times. Relative Weight Rationale: The relationship μ>β>τ>α>γ reflects the priority hierarchy: dynamic obstacle avoidance (highest priority)> static obstacle avoidance > trajectory smoothness > goal-directedness > velocity optimization (lowest priority). This hierarchy ensures safety-first navigation while maintaining efficiency and comfort. Prediction Horizon (T = 3.0 s): The dynamic obstacle prediction time horizon was set to 3.0 seconds based on the robot’s maximum velocity (1.0 m/s) and typical obstacle velocities (0.3–0.8 m/s). This horizon provides sufficient advance warning for collision avoidance while avoiding excessive computational overhead from long-term predictions with high uncertainty. Numerical Stability Constant (λ=10−6): This small positive constant prevents division-by-zero errors in Eq. (20) when the robot’s predicted trajectory passes very close to or intersects with a predicted obstacle position. The value 10−6 is sufficiently small to not affect the cost function behavior while ensuring numerical stability. Table 8 provides a comprehensive summary of all parameters in the enhanced DWA evaluation function. 10.1371/journal.pone.0357770.t008Table 8Complete parameter configuration for enhanced DWA evaluation function (Eq. 21).SymbolDescriptionValueJustification α Heading weight0.5Balances goal-directedness β Static obstacle weight0.8Prioritizes static collision avoidance γ Velocity weight0.3Encourages efficiency τ Historical trajectory weight0.8Ensures trajectory smoothness; adaptive in dynamic scenarios μ Dynamic obstacle weight2.3Highest priority; critical for safety in dynamic environments T Prediction horizon3.0 sBased on robot and obstacle velocities λ Numerical stability constant10−6Prevents division-by-zero errors m Trajectory prediction points30Discrete points over 3.0 s horizonControl frequencyDWA update rate20 HzReal-time performance; 50 ms cycle time Complete parameter configuration for enhanced DWA evaluation function (Eq. 21). 3.2.2.4 Parameter validation and sensitivity analysis: To validate the robustness of the selected parameters, sensitivity analysis was conducted by varying each parameter within ±50% of its nominal value while keeping other parameters fixed. The results indicate: APF parameters (ka, kr, krand): Performance is most sensitive to kr. Reducing kr below 10.0 increases collision rate to 8–12%, while increasing above 20.0 increases path length by 20–30%. The algorithm is relatively robust to variations in ka and krand within ±30%. APF parameters (ka, kr, krand): Performance is most sensitive to kr. Reducing kr below 10.0 increases collision rate to 8–12%, while increasing above 20.0 increases path length by 20–30%. The algorithm is relatively robust to variations in ka and krand within ±30%. DWA basic weights (α, β, γ): Performance is most sensitive to β. Reducing β below 0.6 increases collision risk, while increasing above 1.0 causes overly conservative behavior. The algorithm tolerates ±20% variations in α and γ with less than 5% performance degradation. DWA basic weights (α, β, γ): Performance is most sensitive to β. Reducing β below 0.6 increases collision risk, while increasing above 1.0 causes overly conservative behavior. The algorithm tolerates ±20% variations in α and γ with less than 5% performance degradation. Enhanced DWA weights (τ, μ): Performance is highly sensitive to μ. Reducing μ below 1.5 increases collision rate in dynamic scenarios (8–12%), while increasing above 3.0 causes excessive conservatism. The algorithm is moderately robust to τ variations within ±25%. Enhanced DWA weights (τ, μ): Performance is highly sensitive to μ. Reducing μ below 1.5 increases collision rate in dynamic scenarios (8–12%), while increasing above 3.0 causes excessive conservatism. The algorithm is moderately robust to τ variations within ±25%. Historical trajectory weights (ρ1, ρ2): The algorithm is relatively insensitive to these parameters within ±30%, with less than 8% variation in trajectory smoothness metrics. Historical trajectory weights (ρ1, ρ2): The algorithm is relatively insensitive to these parameters within ±30%, with less than 8% variation in trajectory smoothness metrics. These sensitivity analysis results confirm that the selected parameter values provide a robust operating point that balances multiple performance objectives across diverse scenarios. 3.3 Fusion of ADBI-RRT and DWA The ADBI-RRT algorithm generates a globally suboptimal path with higher motion efficiency compared to the traditional BI-RRT algorithm, owing to its optimized tree expansion strategy and target-biased sampling mechanism. However, its real-time obstacle avoidance performance remains inadequate, particularly in dynamic environments with moving obstacles. In contrast, the improved DWA algorithm exhibits excellent real-time dynamic obstacle avoidance capabilities by virtue of its local motion window sampling and cost function evaluation; nevertheless, it lacks effective global path guidance, which renders it prone to falling into local optima in complex environments. To address these inherent limitations of individual algorithms, this study proposes a novel fusion framework that integrates the ADBI-RRT algorithm with the improved DWA algorithm. This fusion strategy leverages the global path planning advantage of the ADBI-RRT algorithm and the local dynamic obstacle avoidance capability of the improved DWA algorithm, thereby enabling the robot to acquire a globally optimal path while maintaining high-performance real-time obstacle avoidance in dynamic environments. 4. Mathematical simulation and experimental analysis 4.1 ADBI-RRT simulation and fused DWA experiment To rigorously validate the efficacy, computational efficiency, and robust obstacle avoidance performance of the proposed fusion algorithm (ADBI-RRT integrated with enhanced DWA), hereinafter referred to as the ADBI-RRT-DWA hybrid framework, a comprehensive set of numerical simulations was conducted using MATLAB R2024a.All experiments were carried out on a workstation equipped with an AMD Ryzen 7 7435H CPU (3.2 GHz base frequency), NVIDIA RTX 4060 GPU (8 GB VRAM), 16 GB DDR4 RAM (3100 MHz), and a 64-bit Windows 11 Professional operating system. This hardware configuration is consistent with standard setups in robotic path planning research, ensuring the generalizability of the experimental results. It is crucial to clarify the spatial scale representation utilized in this study to ensure internal mathematical consistency. In the MATLAB simulation phase, the environment is modeled using dimensionless normalized coordinate systems to evaluate the geometric and topological performance of the planning algorithms without introducing artificial physical dependencies. The simulated workspaces are specified as dimensionless numerical extents of 50×50, 100×100, and 200×200 (no physical units assigned). All spatial quantities—including obstacle positions (x, y), inter-obstacle distances d(pi,pj), APF influence radii (ρ0=2.5), and path lengths L—are expressed as dimensionless numerical values within this normalized coordinate frame, consistent with the normalized potential field framework established in Section 3.1. When deploying the algorithm to physical robot platforms, these dimensionless coordinates are mapped to the actual physical environment through an explicit scale factor s (in units of m per dimensionless unit). For the ROS-based physical experiments described in Section 5, we adopt s = 1.0 m/unit, such that a 50×50 dimensionless workspace corresponds to a 50×50 m physical area. Critically, the algorithmic parameters remain unchanged across different deployment scales: APF coefficients: ka=0.8, kr=15.0 (dimensionless) APF coefficients: ka=0.8, kr=15.0 (dimensionless) APF influence distance: ρ0=2.5 (dimensionless) APF influence distance: ρ0=2.5 (dimensionless) DWA evaluation weights: α=0.2, β=0.3, γ=0.5 (dimensionless) DWA evaluation weights: α=0.2, β=0.3, γ=0.5 (dimensionless) Only the final velocity commands are converted to physical units (m/s, rad/s) through the robot’s low-level controller. This dimensionless formulation ensures consistent algorithmic behavior while allowing flexible adaptation to different physical environments (e.g., small indoor robots vs. large outdoor vehicles) by simply adjusting the scale factor s without re-tuning parameters. The ADBI-RRT algorithm proposed in this study was compared with four conventional baseline algorithms, namely BI-RRT, RRT*, APF-RRT*, and Informed-RRT*. Comparative experiments were performed in three 2D Cartesian environments with different spatial scales (50×50, 100×100, and 200×200) and gradually increasing obstacle densities (15%, 20%, and 25% obstacle coverage), It is important to clarify the mathematical treatment of the robot’s physical dimensions within this dimensionless framework. The apparent contradiction between treating the robot as a “dimensionless point” while enforcing “non-zero safety margins” is rigorously resolved through standard Configuration Space (C-space) theory (LaValle 2006, Chapter 5), which is fundamental to robot motion planning. This study employs a two-stage hierarchical robot modeling approach: 4.1.1 Stage 1: Global path planning (ADBI-RRT). Prior to the path search phase, all obstacles in the dimensionless workspace are mathematically pre-inflated by a safety distance equal to the robot’s effective radius plus required safety margin. Specifically, for a robot with dimensionless radius rrobot=1.5 (corresponding to 0.15 m physical radius under scale factor s = 0.1 m/unit), each obstacle with original radius robs is inflated to radius robsC−space=robs+rrobot. This obstacle inflation operation transforms the original workspace into a Configuration Space (C-space) representation. Within this inflated C-space, the robot is then rigorously treated as a dimensionless point agent with zero radius. Any collision-free path for this point robot in C-space is mathematically guaranteed to be collision-free for the physical robot (with non-zero radius) in the original workspace. This C-space transformation is a standard technique in sampling-based planners to decouple geometric collision checking from path topology search, significantly improving computational efficiency. 4.1.2 Stage 2: Local trajectory planning (DWA). During real-time trajectory execution, the robot’s actual physical geometry (radius = 0.15 m in physical space, equivalent to 1.5 dimensionless units) and kinematic constraints (maximum linear/angular velocities and accelerations) are explicitly modeled. DWA performs precise collision checking using the robot’s true circular footprint rather than point approximation. For each candidate trajectory in the velocity search space, DWA computes the minimum distance dmin between the robot’s circular boundary and all obstacles, rejecting trajectories where dmin