This paper studies a distributed path-planning problem: How can a sensor network help navigate a nontrial robot to its desired goal in a distributed manner? We consider the case where each sensor node is equipped with sophisticated sensors capable of giving a map for its sensing region. We propose a distributed sampling-based planning algorithm, where every sensor node creates a local roadmap in its locally sensed environment; these local roadmaps are “stitched” together by passing messages among nodes and forming a larger implicit roadmap without having a global representation. Based on the implicit roadmap, a feasible path is computed in a distributed manner, and the robot moves along the path by interacting with sensor nodes, each of which giving a portion of the path within the local environment of the node. Simulations show that the algorithm is able to solve the path-planning problem with low communication overhead.

This paper studies a distributed path planning problem: how can a sensor network help to navigate a robot to its desired goal in a distributed manner. We consider the case where each sensor node is equipped with sophisticated sensors capable of giving a map for its sensing region. We propose a distributed sampling based planning framework (Distributed PRM), where every sensor node creates a local roadmap in its locally-sensed environment; these local roadmaps are “stitched” together by passing messages among nodes and form a larger implicit roadmap without having a global representation. Based on the implicit roadmap, a feasible path is computed in a distributed manner, and the robot moves along the path by interacting with sensor nodes, each of which gives a portion of the path within the local environment of the node. Preliminary simulations show the proposed framework is able to solve path planning problem with low communication overhead.

While a network of autonomous mobile agents is capable of performing spatially distributed tasks, communication between agents imposes a class of constraints over the corresponding task. This paper proposes a distributed paradigm to deal with a critical communication constraint, connectedness constraint, where a group of mobile agents are required to remain connected while performing a task (e.g., formation control, consensus, etc.). The proposed method adaptively extracts communication backbone of the group, which is formed by a subset of agents, and thus partitions the group into backbone agents and non-backbone agents. The connectedness of the system is maintained at two levels: motion of backbone agents is controlled to maintain existing connections in the backbone; motion of non-backbone agents is determined via a leader-follower formation control method with backbone agents as the leaders. Key advantages of the proposed approach are that it can deal with arbitrary system topologies, it is a distributed method, it uses only two-hop neighbor information, and has low communication cost.

In this paper, we propose a distributed path planning method for robot navigation amidst a wireless sensor network. Our method uses communication backbone as a roadmap. In building and maintaining the roadmap, it takes path safety and network longevity into account, and therefore the roadmap adapts to dynamic dangers and evolves over time to increase network longevity. To find a safe path, a navigation field is propagated over the roadmap and the shortest path is computed. Simulation results show that as compared to existing methods our method finds a safe path with less communication cost, and in dense networks it generates smaller roadmaps. We also provide theoretical bounds on the path quality in terms of path length.

We consider the problem of planning collision-free paths for a redundant robot manipulator whose end-effector must travel along a specified path. A probabilistic method has been proposed for the problem, which does not allow self-motions of the robot as it moves along the end-effector path. In this paper, we propose an enhancement, which allows such self-motions. This is primarily accomplished by explicitly representing self-motions for a certain pose as a self-motion graph, which is explored with probabilistic techniques for closed-chain robots. Computer simulations show that this enhancement improves performance in most cases. Depending on the limits set on the run-time (always needed in practice for probabilistic sampling methods), the planner with self-motion enhancement will find a path where the original algorithm without self-motion may not

In this paper, we address the path planning problem with general end-effector constraints (PPGEC) for robot manipulators. Two approaches are proposed. The first approach is adapted from an existing randomized gradient descent (RGD) method for closed-chain robots. The second approach is radically different. We call it ATACE alternate task-space and configuration-space exploration. Unlike the first approach which searches purely in C-space, ATACE works in both task space and C-space. It explores the task space for end-effector paths satisfying given constraints, and utilizes trajectory tracking technique(s) as a local planner(s) to track these paths in the configuration space. We have implemented both approaches and compare their relative performances in different scenarios. ATACE outperforms RGD in majority (but not all) of the scenarios. We outline intuitive explanations for the relative performances of these two approaches.