Analysis of a Spatialized Brain-Body-Environment System
Denizhan Pak, Quan Le Thien, Christopher J. Agostino · 2025
Read Analysis of a Spatialized Brain-Body-Environment System by Denizhan Pak, Quan Le Thien, Christopher J. Agostino (2025) in the Tower of Babel, a free online library of public-domain works preserved in clean, searchable text. No sign-up. The full text is served as a plain-text file at a predictable URL, where text is present: https://antarctic-labs.com/tower-of-babel/arxiv-2509-25640.txt.
Introduction
Denizhan Pak, Quan Le Thien, and Christopher Agostino rebuild a brain-body-environment model so that space itself plays a role, in a paper from 2025. Such models study how an agent's brain, body, and world shape each other. Most treat the brain as a web of point equations with no spread. That choice hides how signals travel across tissue.
Their version uses partial differential equations for both brain and body. As a test case, they revisit a model of a child on a swing. They vary where input lands and how it spreads along the tissue. Those spatial choices change the motion that comes out. The authors argue that place and travel time shape adaptive acts. The frame ties abstract neural models closer to real bodies.
Catalog record
| CREATOR | Denizhan Pak, Quan Le Thien, Christopher J. Agostino |
|---|---|
| YEAR | 2025 |
| COLLECTION | PAPERS |
| FORMAT | TXT |
| FILE SIZE | 36 KB |
| VERSION | 1.0 |
| LICENSE | http://creativecommons.org/licenses/by/4.0/ |
| RIGHTS | CC BY |
| SOURCE | arXiv |
Source: https://arxiv.org/abs/2509-25640
Browse the full library · About the Tower of Babel
Full text
The opening of the work, from the library's hosted text:
Analysis of a Spatialized Brain-Body-Environment System Denizhan Pak, Quan Le Thien, Christopher J. Agostino Abstract The brain-body-environment framework studies adaptive behavior through embodied and situated agents, emphasizing interactions between brains, biomechanics, and environmental dynamics. However, many models often treat the brain as a network of coupled ordinary differential equations (ODEs), neglecting finer spatial properties which can not only increase model complexity but also constrain observable neural dynamics. To address this limitation, we propose a spatially extended approach using partial differential equations (PDEs) for both the brain and body. As a case study, we revisit a previously developed model of a child swinging, now incorporating spatial dynamics. By considering the spatio-temporal properties of the brain and body, we analyze how input location and propagation along a PDE influence behavior. This approach offers new insights into the role of spatial organization in adaptive behavior, bridging the gap between abstract neural models and the physical constraints of embodied systems. Our results highlight the importance of spatial dynamics in understanding brain-body-environment interactions. Introduction Adaptive behavior is often modeled using a brain-body-environment framework ( Chiel and Beer, 1997 ) . Within this framework, the body and environment are often modeled as a mechanical systems. Neural dynamics couple with this mechanical system as a collection of ordinary differential equations (ODEs), often implemented as a recurrent neural network ( Beer, 2003 ) . This approach has been foundational to the study of adaptive behavior and has yielded many experimentally validated insights ( Beer, 2000 ) . However, this paradigm faces some theoretical limitations in its ability to model some of the geometric relations that can shape neural dynamics. Spatial properties of neural networks are reduced to edge weights between nodes, obscuring the spatial extent of the nodes themselves and the complex geometry observed in real neural tissue ( Paik et al., 2020 ) . Additionally, while external time delays can be added to ODEs, PDEs provide a natural trade-off between space and time that demonstrates delay dynamics. Neural field models, which emphasize spatial dynamics, have a long history in neuroscience ( Coombes et al., 2014 ) . However, research integrating neural fields with biomechanical simulations remains underexplored ( Dale and Husbands, 2010 ) . Partial differential equations (PDEs) are natural tools for neuromechanical modeling, as they capture both temporal and spatial dynamics critical to understanding adaptive behavior within brain-body-environment frameworks. PDEs offer a different perspective from ODE-based neural mass network models ( Pinotsis et al., 2013 ) . Their vector fields depend on both time and space, enabling them to incorporate spatial constraints imposed by embodied interactions with the environment. Additionally, PDEs can model signal propagation more accurately: while ODEs assume instantaneous signal transmission across nodes, PDEs explicitly represent spatial signal propagation through a medium. Therefore, they can be used to model delays in signaling, gradients of activity and neural wave patterns. The cost of this expressivity is increased complexity. However, simplifying assumptions—as we demonstrate—can mitigate this complexity and enable analytical tractability. We explore PDE-based neuromechanical modeling through the lens of minimal cognition ( Beer, 1995 ) . Minimal cognitive modeling studies adaptive behavior in simplified brain-body-environment systems performing tasks that serve as conceptual analogs to real-world complexity ( Beer, 2003 ) . This approach has proven effective for clarifying philosophical interpretations ( Varela et al., 1991 ; Clark, 1997 ) , developing new tools for model creation and analysis ( Beer, 1995 ) , and inspiring experimental applications ( Marder, 2020 ) . For this work, we adopt the reactive swinging agent system introduced by Thorniley and Husbands ( Thorniley and Husbands, 2013 ). This model simulates a child swinging on a swing, aligning with minimal cognitive modeling due to its simplicity in biomechanical and environmental components. The system also features well-defined optimal behavior within specific parameter ranges, facilitating the conversion of its original ODE-based neural controller into a PDE framework. The remainder of this paper is structured as follows: Section 2 details Thorniley and Husbands’ original model, including a bifurcation analysis of the sensory-coupling parameter. Section 3 describes the reformulation of the agent’s state variables as PDEs, highlighting dynamical differences from the original model. Section 4 introduces a simplifying assumption to analyze these differences. Section 5 explores the implications of this assumption. Finally, Section 6 examines signal propagation through spatialized sensors and effectors, demonstrating how spatial parameters influence task performance. A Simple Brain-Body-Environment System Figure 1: The neuromechanical model originally presented in Thorniley and Husbands, 2013 . The agent is represented as a mass-spring which is attached to then end of a rigid rod. The rod can rotate about the top and \theta denotes the deviation from the rod in its steady resting state. The dynamics of the movement of the rod are governed by the gravitational and torsional forces which are affected by the length of the agent r . The agent can then actuate the value of r to move the rod (swing). Our system of interest is the reactive swinging agent proposed by Thorniley and Husbands, 2013 . The swing is modeled as a rigid, massless rod attached to a fixed pivot at one end, with a mass (representing the agent) at the other end. We visualize this system in figure 1 . The original system is governed by the following equations: \displaystyle\dot{\theta}=\omega (1) \displaystyle\dot{\omega}=-\frac{g}{1+r}\sin(\theta)-b\omega (2) \displaystyle\dot{r}=v (3) \displaystyle\dot{v}=\frac{u}{1+r^{2}}+g\cos(\theta)+(1+r)\omega^{2}-kr-cv (4) \displaystyle\dot{u}=\phi(A\tanh(\rho v)-u) (5) This 5-dimensional system of ordinary differential equations (ODEs) separates into environmental states (\theta,\omega) and agent states (r,v,u) . Here, \theta is the swing angle, \omega the angular velocity, r the pendulum extension controlled by the agent, v the extension rate, and u the neural force acting on v,r . Descriptions and numerical values can for those and other parameters can be found in Table 1 . Symbol Type/(Initial Value/Range) Description \theta Variable ( \pi ) Angle of pendulum from downward vertical \omega Variable ( 0.1 ) Angular velocity of pendulum r Variable ( 0.1 ) Current pendulum extension v Variable ( 0.1 ) Rate of pendulum extension u Variable ( 0.1 ) State variable of agents nervous system A Parameter (0,300) Strength of sensorimotor coupling g 9.81 Acceleration due to gravity b 0.3 Pendulum damping coefficient \rho 2 Motor neuron sensitivity \phi 20 Control parameter k 100 Spring force constant c 20 Spring damping (= 2\sqrt{k} for critical damping) x_{s} Parameter (-1,1) Central spatial position on v that is driven by r \sigma_{s} Parameter (0,\infty) Standard deviation of the spatial part on v that is driven by r x_{e} Parameter (-1,1) Spatial location v(x_{e},t) that drives by r (x_{\sf min},x_{\sf max}) Parameter (-1,1) Boundary points of v and u Table 1: Table of variables and parameters, adapted from Thorniley and Husbands, 2013 . The numbers next to the variables represent their initial value used for the simulations. The numbers next to parameters represent their ranges across plots. To analyze the system, we varied the parameter A , which governs the strength of sensory coupling strength between u and v . Previously, Thorniley and Husbands, 2013 identified three dynamical regimes as A increases: 1. Dampened regime: In this regime, the system dampens to a fixed point neither the environment nor the agent show any interesting long-term behavior. This can be