Reactive Baseline Framework for Dynamic Navigation and Obstacle Avoidance of Bipedal Robots in Simulated Indoor Environments

Abstract

Autonomous navigation in dynamic indoor environments remains a significant challenge for bipedal robots due to the need to maintain locomotion stability while responding to moving obstacles. This paper presents a lightweight reactive baseline framework for biped robots developed in ROS Noetic and Gazebo simulation. The proposed framework, inspired by existing reactive avoidance techniques for bipedal platforms, integrates a finite-state machine (FSM) navigation controller with a simple lateral sidestepping strategy for dynamic obstacle avoidance, enabling continuous goal-directed motion without complex perception or prediction modules. A custom URDF(Unified Robot Discription Format)-based biped robot model and a sinusoidally moving obstacle were used for evaluation. One hundred continuous experimental trials were conducted under identical conditions. The framework achieved a 95.0% success rate in completing the full dynamic navigation task (goal reaching, obstacle avoidance, and return-to-home). The 5.0% failure rate was explicitly analyzed and attributed to reactive oscillation and the lack of upper-torso angular momentum regulation during high-velocity lateral maneuvers. For successful runs, the framework demonstrated a highly stable average task completion time of approximately 11 seconds. These results validate the effectiveness of the proposed lightweight reactive approach as a strong, predictable baseline for future research in biped robot navigation. These results validate the effectiveness of the proposed lightweight reactive approach as a strong baseline for future research in biped robot navigation, particularly in resource-constrained or perception-limited settings.

Introduction

The text presents a lightweight navigation framework for biped robots operating in dynamic indoor environments. Biped robots are useful because they can move through human-oriented spaces, narrow passages, and areas with obstacles, but navigation is challenging because the robot must maintain balance while responding to changing surroundings.

The study focuses on a simple and computationally efficient solution rather than complex perception, optimization, or learning-based methods. It combines a finite-state navigation controller with a reactive obstacle-avoidance strategy and evaluates the approach in a ROS Noetic–Gazebo simulation environment.

Main Objectives

• Develop a modular ROS-Gazebo framework for biped robot navigation.

• Implement a finite-state navigation controller for goal-directed movement.

• Add a lightweight reactive mechanism for avoiding dynamic obstacles.

• Evaluate the robot's navigation performance under moving-obstacle conditions.

• Establish a baseline for future research involving advanced perception, locomotion, and navigation techniques.

Literature Review

Previous research has explored several approaches to biped navigation, including:

• Reactive methods, such as Dynamic Window and virtual-force approaches.

• Deep reinforcement learning for autonomous walking and obstacle avoidance.

• Human-inspired navigation for shared environments.

• Visual SLAM and dynamic-object detection for perception.

• Optimization-based whole-body and footstep planning.

• Control Lyapunov and Control Barrier Functions for safety.

• Deep-learning-based obstacle avoidance.

While these approaches can provide sophisticated navigation capabilities, many require substantial computational resources, complex models, or extensive parameter tuning. The present work instead emphasizes simplicity, reproducibility, and low computational cost.

Proposed System

The framework consists of four major components:

1. Biped robot model – A simplified humanoid with pelvis, thigh, shin, and foot segments. Hip, knee, and ankle joints are modeled using effort-based control.

2. Indoor simulation environment – Contains walls, corridors, and a moving obstacle.

3. Reactive navigation controller – Uses a finite-state machine to control movement toward the goal and back home.

4. Dynamic obstacle module – Simulates a moving object using sinusoidal motion.

Robot Model and Simulation

The biped model is created using URDF and simulated in Gazebo. Foot contact sensors are included to improve ground interaction and simulation stability. The robot is intentionally simplified so that the a finite-state controller with reactive sidestepping, the framework allows a simulated biped robot to move toward a goal while study can focus on navigation and obstacle avoidance rather than detailed humanoid gait optimization.

The moving obstacle follows a sinusoidal trajectory:

• Amplitude: 0.6 m

• Angular frequency: 0.5 rad/s

• Approximate period: 12.6 seconds

This produces repeated interactions between the robot and the dynamic obstacle.

Navigation Controller

The controller uses four states:

• HOME – initialization state.

• GO TO GOAL – robot moves toward the destination.

• RETURN HOME – robot returns after reaching the goal.

• DONE – navigation terminates.

The robot moves incrementally toward its target. The distance between the robot and obstacle is continuously calculated. When the obstacle comes within a 1.0 m safety threshold, the avoidance mechanism is activated.

Obstacle Avoidance

The proposed strategy uses an immediate lateral sidestep instead of prediction or optimization.

• Sidestep distance: 0.18 m

• Avoidance direction alternates between right and left.

• Alternating directions helps reduce cumulative sideways drift during repeated obstacle encounters.

This makes the method computationally lightweight and relatively easy to implement and reproduce.

Conclusion

This paper presented a lightweight reactive baseline framework for bipedal robot navigation in dynamic indoor environments. The proposed methodology integrated a custom URDF-based biped model, a finite-state machine navigation controller, and a lateral sidestepping strategy for reactive obstacle avoidance. One hundred continuous automated trials yielded a 95.0% success rate, successfully demonstrating collision-free dynamic navigation with a stable average task completion time of approximately 11 seconds. The 5.0% failure rate distinctly identified the physical limitations of executing rapid oscillatory corrections without upper-body angular momentum regulation. These results demonstrate that the proposed lightweight reactive approach is highly effective, providing a robust and reproducible baseline for future research in bipedal robot navigation. The quantitative experimental results, supported by the six Gazebo simulation visual sequences (Figures 3a–3f), confirm that the integration of a finite-state controller with reactive sidestepping effectively manages goal-directed navigation and dynamic obstacle avoidance. Notably, this is achieved without reliance on computationally expensive perception modules or global planning algorithms. While the current framework demonstrates reliable performance, several structural enhancements are planned. Future work will focus on: 1) Integrating Adaptive Monte Carlo Localization (AMCL) or visual SLAM to enable navigation in unstructured, unknown environments. 2) Implementing dynamic obstacle tracking and trajectory prediction to optimize avoidance maneuvers. 3) Developing advanced humanoid footstep planning and full-body locomotion control algorithms. 4) Validating the proposed framework and subsequent extensions on a physical bipedal hardware platform. The modular architecture of the proposed system (Figure 1) facilitates the seamless integration of these future extensions. Ultimately, this baseline framework is intended to serve as a practical reference architecture for the bipedal robotics research community across both simulated and real-world domains.

References

[1] A. K. Kashyap and D. Parhi, “Human-inspired dynamic obstacle and inter-collision avoidance algorithm for humanoid biped robots,” Robotics and Autonomous Systems, vol. 169, p. 105023, 2025. [2] H. Sugiura, M. Gienger, H. Janssen, and C. Goerick, “Reactive self collision avoidance with dynamic task prioritization for humanoid robots,” International Journal of Humanoid Robotics, vol. 7, no. 1, pp. 1–24, 2010. [3] L. Rossini, E. M. Hoffman, S. H. Bang, L. Sentis, and N. G. Tsagarakis, “A real-time approach for humanoid robot walking including dynamic obstacles avoidance,” in IEEE-RAS International Conference on Humanoid Robots, 2023. [4] R. P. Singh, M. Benallegue, M. Morisawa, R. Cisneros, and F. Kanehiro, “Learning bipedal walking on planned footsteps for humanoid robots,” arXiv preprint arXiv:2207.12644, 2022. [5] J. Guzzi, A. Giusti, L. M. Gambardella, G. Theraulaz, and G. A. Di Caro, “Human-friendly robot navigation in dynamic environments,” in Proceedings of the IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS), 2018. [6] M. R. U. Saputra, A. Markham, and N. Trigoni, “Visual slam and structure from motion in dynamic environments: A survey,” ACM Computing Surveys, vol. 51, no. 2, p. 37, 2018. [7] C. Yu, Z. Liu, X.-J. Liu, F. Xie, Y. Yang, Q. Wei, and Q. Fei, “Ds-slam: A semantic visual slam towards dynamic environments,” arXiv preprint arXiv:1805.09144, 2018. [8] S. Kuindersma, R. Deits, M. Fallon, A. Valenzuela, H. Dai, F. Permenter, T. Koolen, P. Marion, and R. Tedrake, “Optimization-based locomotion planning, estimation, and control design for the atlas humanoid robot,” Autonomous Robots, vol. 40, no. 3, pp. 429–455, 2016. [9] J.-K. Huang and J. W. Grizzle, “Efficient anytime clf reactive planning system for a bipedal robot on undulating terrain,” arXiv preprint arXiv:2106.06693, 2021. [10] J. Liu, M. Li, J. W. Grizzle, and J.-K. Huang, “Realtime safety control for bipedal robots to avoid multiple obstacles via clf-cbf constraints,” arXiv preprint arXiv:2301.01906, 2023. [11] R. Ba, P. Mishra, M. Kalaimani, P. Juyal, P. K. Anjani, R. Dua, and P. Mamoria, “Real-time obstacle avoidance in mobile robots using deep reinforcement learning,” Future Technology Journal, vol. 5, no. 3, pp. 69–76, 2026. [12] K. Wang, H. Fei, and P. Kormushev, “Fast online optimization for terrain-blind bipedal robot walking with a decoupled actuated slip model,” Frontiers in Robotics and AI, vol. 9, p. 812258, 2022. [13] M. Labbe and F. Michaud, “Rtab-map as an open-source lidar and visual slam library for large-scale and long-term online operation,” Journal of Field Robotics, 2019. [14] Z. Zhang, H. Hajieghrany, E. Dean et al., “Prescient collision-free navigation of mobile robots with iterative multimodal motion prediction of dynamic obstacles,” IEEE Robotics and Automation Letters, vol. 8, no. 9, pp. 5488–5495, 2023.

Copyright

Copyright © 2026 Jonnavada Ravi Kiran, Akepogu Ananda Rao, Palagiri Radhika Raju. This is an open access article distributed under the Creative Commons Attribution License, which permits unrestricted use, distribution, and reproduction in any medium, provided the original work is properly cited.