Active Modular Environment for Robot Navigation

Shota Kameyama, Keisuke Okumura, Yasumasa Tamura, Xavier Défago

Abstract

This paper presents a novel robot-environment interaction in navigation tasks such that robots have neither a representation of their working space nor planning function, instead, an active environment takes charge of these aspects. This is realized by spatially deploying computing units, called cells, and making cells manage traffic in their respective physical region. Different from stigmegic approaches, cells interact with each other to manage environmental information and to construct instructions on how robots move. As a proof-of-concept, we present an architecture called AFADA and its prototype, consisting of modular cells and robots moving on the cells. The instructions from cells are based on a distributed routing algorithm and a reservation protocol. We demonstrate that AFADA achieves efficient robot moves for single-robot navigation in a dynamic environment changing its topology with a stochastic model, comparing to self-navigation by a robot itself. This is followed by several demos, including multi-robot navigation, highlighting the power of offloading both representation and planning from robots to the environment. We expect that the concept of AFADA contributes to developing the infrastructure for multiple robots because it can engage online and lifelong planning and execution.

Introduction

Traditional robot navigation typically relies on a "selfish" paradigm: each agent is equipped with complex sensors to build internal maps (SLAM) and calculate paths independently. This approach introduces significant computational burdens, struggles with real-time updates in dynamic environments, and requires complex coordination to prevent multi-robot collisions.

To address these challenges, we propose a paradigm shift inspired by spatial computing. Instead of the robot modeling the world, the environment models itself and directs the robot. This is realized through AFADA (Adaptive Fully Automated Decentralized Architecture), where the "brains" of navigation are shifted from the mobile agent to the floor infrastructure.

System Architecture

The AFADA system is composed of two primary modular components designed to offload computational complexity from the mobile agents to the physical workspace.

  • Active Cells: These are modular, square computing units based on the ESP32 microcontroller. They form a physical grid and communicate with immediate neighbors via serial connections. Interaction with robots is handled through Near Field Communication (NFC).
  • Simple Robots: Built on a Zumo 32U4 base, these robots lack global maps or high-level planning functions. They act as simple actuators that follow step-by-step instructions received from the specific cell they currently occupy.
The AFADA prototype system, consisting of modular computing cells and simple mobile robots.

Coordination and Routing

Navigation is managed through two key decentralized protocols:

  1. Self-Stabilizing Routing: Cells maintain local routing tables similar to internet data packets. If a cell is added, removed, or fails, neighbors detect the change and propagate updated distance information across the grid until a valid state is reached.
  2. Reservation Protocol: To ensure collision-free movement, a robot must "reserve" its next target cell. The environment acts as a distributed traffic controller, only granting permission to move if the target cell is unoccupied.

Experimental Evaluation

To test the resilience of AFADA, we evaluated it against a "self-navigating" baseline in custom experimental fields (simple-loop, two-bridge, and two-loop). These fields were subjected to stochastic failure conditions where cells would become impassable and recover with specific probabilities ($p$ and $q$), simulating a highly dynamic topology.

Results and Discussion

The experimental results highlight the efficiency of offloading planning to the environment. In the "self-nav" baseline, robots only replan upon physically encountering an obstacle. In contrast, AFADA's active cells update the global path before the robot reaches a blockage.

Navigation efficiency (steps) comparison between AFADA and a self-navigating baseline in dynamic environments.

Key findings include:

  • Efficiency in Dynamic Settings: In highly unstable environments (low recovery probability $q=0.01$), AFADA was significantly more efficient. In the two-loop field, AFADA achieved a median of 44 steps compared to 50 for the self-navigating robot ($P < 0.01$).
  • Multi-Robot Coordination: Demos such as automated parking showed that the environment could guide multiple robots to vacant spots and manage one-way traffic without any direct robot-to-robot communication.
  • Resilience: The system demonstrated real-time rerouting capabilities as cells were physically added or removed during active operation.

Conclusion

AFADA demonstrates that embedding intelligence into the physical infrastructure can drastically simplify robot design and improve navigation efficiency in dynamic, multi-agent spaces. While the current prototype faces challenges regarding deployment costs and the scale of robots relative to cell size, it provides a robust framework for lifelong planning and execution in robotic infrastructure. Future work will focus on more robust self-stabilizing protocols and coordination logic for larger agents.

Summary updated

Abstract

This paper presents a novel robot-environment interaction in navigation tasks such that robots have neither a representation of their working space nor planning function, instead, an active environment takes charge of these aspects. This is realized by spatially deploying computing units, called cells, and making cells manage traffic in their respective physical region. Different from stigmegic approaches, cells interact with each other to manage environmental information and to construct instructions on how robots move. As a proof-of-concept, we present an architecture called AFADA and its prototype, consisting of modular cells and robots moving on the cells. The instructions from cells are based on a distributed routing algorithm and a reservation protocol. We demonstrate that AFADA achieves efficient robot moves for single-robot navigation in a dynamic environment changing its topology with a stochastic model, comparing to self-navigation by a robot itself. This is followed by several demos, including multi-robot navigation, highlighting the power of offloading both representation and planning from robots to the environment. We expect that the concept of AFADA contributes to developing the infrastructure for multiple robots because it can engage online and lifelong planning and execution.
BibTeX
@inproceedings{kameyama2021active,
  author = {Shota Kameyama and Keisuke Okumura and Yasumasa Tamura and Xavier Défago},
  title = {{Active Modular Environment for Robot Navigation}},
  booktitle = {IEEE International Conference on Robotics and Automation},
  year = {2021},
  doi = {10.1109/icra48506.2021.9561111},
}
Figure