1. What is Multi-Robot Coverage Path Planning (MCPP)?
Multi-Robot Coverage Path Planning (MCPP) is an extension of Coverage Path Planning (CPP) that involves coordinating the paths of multiple robots to achieve complete coverage of a given terrain. It improves coverage task efficiency and system robustness, playing a crucial role in applications such as search and rescue, environmental monitoring, and mapping. The main challenges in MCPP include distributing robots effectively while avoiding collisions, ensuring complete coverage, and minimizing coverage time. Offline planning is commonly used to solve MCPP without inter-robot communication. A Mixed Integer Programming (MIP) model is proposed to optimally solve Min-Max Rooted Tree Cover (MM-RTC), with two efficient suboptimal heuristics designed to reduce model size while maintaining completeness. The MIP-based MCPP planner yields higher-quality solutions but at the cost of longer runtime.
read more
2. What are the main categories of MCPP algorithms?
MCPP algorithms can be categorized into two main categories: decomposition-based and graph-based methods. Decomposition-based methods partition the terrain geometrically and generate zigzag coverage paths within each region. However, they are not suitable for weighted terrains with nonuniform traversal costs and obstacle-rich terrains like mazes. On the other hand, graph-based methods operate on a graph representation of the terrain, considering varying traversal costs and providing more flexibility. These methods have been discussed in detail in the provided section.
read more
3. What is the graph-based Spanning Tree Coverage (STC) used for in the context of terrain coverage?
Graph-based Spanning Tree Coverage (STC) is used for both weighted and unweighted terrains to address the coverage problem. It operates on a 2D 4-neighbor grid graph G = (V, E) that represents the terrain to be covered. STC and its multi-robot variants generate coverage paths on a decomposition D = (V d , E d ) of G = (V, E). Each vertex v V is decomposed into four smaller adjacent vertices, and each resulting vertex u V d is assigned a weight w(u) = w(v)/4 and must be visited by at least one robot for complete coverage. The edges (u, v) in both E and E d are connected for each pair of vertically or horizontally adjacent vertices and are assigned a weight [w(u) + w(v)] /2. The coverage time t(p) of a valid complete coverage path p = (v 1 , v 2 , ..., v |p| ) in D is defined as t(p) = |p| i=1 [(w(v i ) + w(v i+1 )) /2], where v 1 = v |p|+1 is the robot's given initial vertex from V d. STC is time-optimal using any spanning tree of G in unweighted terrain and using the minimum spanning tree of G in weighted terrain. It has also been extended for Multi-Robot Path Planning Problem (MCPP).
read more
4. What is the formulation of the MIP model for MMRTC?
The MIP model for MMRTC is formulated as follows: minimize x,y,f,t t EQUATION EQUATION y i ri = 1, i I (6) vV y i v = 1 + eE x i e , i I (7) f i e,u + f i e,v = x i e , e = (u, v) E, i I (8) eE e~v f i e,v <= 1 - 1 |V | , v V, i I (9) x i e , y i v {0, 1}, v V, e E, i I (10) f i e,u , f i e,v , t R + , e = (u, v) E, i I (11) The model includes binary variables x and y, continuous flow variables f, and the makespan variable t. The constraints ensure that each vertex is included in at least one subtree, each subtree is a single tree, and there are no cycles in the subtrees. The objective function minimizes the maximum weight among all subtrees.
read more