Multi-robot path planning for budgeted active perception with self-organising maps

Abstract

Abstract-We propose a self-organising map (SOM) algorithm as a solution to a new multi-goal path planning problem for active perception and data collection tasks. We optimise paths for a multi-robot team that aims to maximally observe a set of nodes in the environment. The selected nodes are observed by visiting associated viewpoint regions defined by a sensor model. The key problem characteristics are that the viewpoint regions are overlapping polygonal continuous regions, each node has an observation reward, and the robots are constrained by travel budgets. The SOM algorithm jointly selects and allocates nodes to the robots and finds favourable sequences of sensing locations. The algorithm has polynomial-bounded runtime independent of the number of robots. We demonstrate feasibility for the active perception task of observing a set of 3D objects. The viewpoint regions consider sensing ranges and selfocclusions, and the rewards are measured as discriminability in the ensemble of shape functions feature space. Simulations were performed using a 3D point cloud dataset from a real robot in a large outdoor environment. Our results show the proposed methods enable multi-robot planning for budgeted active perception tasks with continuous sets of candidate viewpoints and long planning horizons.

Decentralised Monte Carlo Tree Search for Active Perception

“…Efficient non-myopic solutions can be designed by exploiting problem-specific characteristics [4,6]. But in general, the problem is a POMDP, which is notoriously difficult to solve.…”

“…The problem is motivated by tasks where a team of Dubins robots maximally observes a set of features of interest in an environment, given a travel budget [4]. Each feature can be viewed from multiple viewpoints and each viewpoint may be in observation range of multiple features.…”

“…Each feature can be viewed from multiple viewpoints and each viewpoint may be in observation range of multiple features. This formulation generalises the orienteering problem [12] by combining the set structure of the generalised travelling salesman problem with the budget constraints of the orienteering problem with neighbourhoods [9] extended for multi-agent scenarios [4].…”

“…For the objective function, we have a collection of sets S = (S 1 , S 2 , ...), where each S i ⊆ V. These sets may represent a set of features of interest, where a vertex is an element of a set only if the associated feature can be observed from the vertex location. We assume each set is a disc, however the formulation could extend to more complex models [4]. The vertices v j ∈ V are randomly placed within the sets.…”

Online planning for multi-robot active perception with self-organising maps

“…A preliminary version of this paper appeared as Best et al (2016b). This extended version additionally contains: expanded algorithmic details; a generalised formulation for fixed start and/or end locations (similar to ); extended theoretical analysis of the runtime complexity and convergence; empirical analysis of the algorithm's convergence and anytime properties; empirical validation of an example observation model definition; a comparison to Dec-MCTS; a formulation for online scenarios; and extensive simulation experiments that demonstrate the feasibility of the approach for online replanning scenarios.…”