Robot Navigation in Unknown and Cluttered Workspace with Dynamical System Modulation in Starshaped Roadmap

TL;DR

This work tackles navigation in completely unknown, cluttered environments by introducing a dynamic starshaped roadmap that represents free space with starshaped regions generated from real-time point-cloud data. A high-frequency DSM-based reactive controller modulates motion inside and across starshaped regions, enabling smooth, safe navigation even in complex obstacle configurations. The approach uses frontier-point exploration to incrementally construct a connected roadmap, handles dead-ends with updates, and weighs overlaps to steer motion through overlapping regions. Extensive simulations and real-world experiments show higher success rates and reduced travel times compared with state-of-the-art baselines, demonstrating the method’s robustness and practical applicability. The overall contribution is a real-time, perception-driven framework that efficiently leverages starshaped geometry for autonomous navigation in unknown, cluttered spaces.

Abstract

Compared to conventional decomposition methods that use ellipses or polygons to represent free space, starshaped representation can better capture the natural distribution of sensor data, thereby exploiting a larger portion of traversable space. This paper introduces a novel motion planning and control framework for navigating robots in unknown and cluttered environments using a dynamically constructed starshaped roadmap. Our approach generates a starshaped representation of the surrounding free space from real-time sensor data using piece-wise polynomials. Additionally, an incremental roadmap maintaining the connectivity information is constructed, and a searching algorithm efficiently selects short-term goals on this roadmap. Importantly, this framework addresses dead-end situations with a graph updating mechanism. To ensure safe and efficient movement within the starshaped roadmap, we propose a reactive controller based on Dynamic System Modulation (DSM). This controller facilitates smooth motion within starshaped regions and their intersections, avoiding conservative and short-sighted behaviors and allowing the system to handle intricate obstacle configurations in unknown and cluttered environments. Comprehensive evaluations in both simulations and real-world experiments show that the proposed method achieves higher success rates and reduced travel times compared to other methods. It effectively manages intricate obstacle configurations, avoiding conservative and myopic behaviors.

Figures (6)

• Figure 1: Highlight of the proposed navigation framework. (a) Our method, a dynamic starshaped roadmap is constructed in real-time to facilitate the proposed reactive controller based on DSM, enabling the robot to navigate smoothly towards its goal. (b) With our approach, the robot successfully navigates through unknown and cluttered environments, demonstrating its effectiveness in real-world scenarios. (c)-(e) Comparing with the representation of free regions by ellipses and convex polygons, our approach uses starshaped regions, covering a larger area and enabling better use of perception information for navigation.
• Figure 2: Illustration of the navigation process on our dynamically constructed starshaped roadmap. (a) The roadmap is initialized with the starshaped region at the start position. (b) Frontier points are generated on the roadmap. (c) The short-term goal is subsequently selected based on the search algorithm, and a new starshaped traversable region is extracted upon reaching the current short-term goal if it is extendable. (d) The navigation process is carried on with repeatedly calculated short-term goals till the target is reached.
• Figure 3: Ablation study of the proposed adaptive segmentation method for piecewise polynomials. Compared to different fixed segment numbers, our method achieves higher accuracy while offering a trade-off in computation time.
• Figure 4: In the simulation experiments, the start and target positions will be randomly generated in the corresponding red and green regions.
• Figure 5: The simulation results, with methods from left to right: DSM, FOA, FOA combined with RM, and our approach, demonstrate that our method effectively handles both environments.