Robustness-driven Exploration with Probabilistic Metric Temporal Logic

Abstract

The ability to perform autonomous exploration is essential for unmanned aerial vehicles (UAV) operating in unstructured or unknown environments where it is hard or even impossible to describe the environment beforehand. However, algorithms for autonomous exploration often focus on optimizing time and coverage in a greedy fashion. That type of exploration can collect irrelevant data and wastes time navigating areas with no important information. In this paper, we propose a method for exploiting the discovered knowledge about the environment while exploring it by relying on a theory of robustness based on Probabilistic Metric Temporal Logic (P-MTL) as applied to offline verification and online control of hybrid systems. By maximizing the satisfaction of the predefined P-MTL specifications of the exploration problem, the robustness values guide the UAV towards areas with more interesting information to gain. We use Markov Chain Monte Carlo to solve the P-MTL constraints. We demonstrate the effectiveness of the proposed approach by simulating autonomous exploration over Amazonian rainforest where our approach is used to detect areas occupied by illegal Artisanal Small-scale Gold Mining (ASGM) activities. The results show that our approach outperform a greedy exploration approach (Autonomous Exploration Planner) by 38% in terms of ASGM coverage. 1 .

Deep Reinforcement Learning for Adaptive Exploration of Unknown Environments

“…Robustness of Metric Temporal Logic (MTL) has been embedded in A* to increase the safety of UAVs navigating adversarial environments [19]. In [3], the NBV is sampled according to the current field of view of the UAV. Here, the views are randomly sampled as potential targets via Markov Chain Monte Carlo methods and evaluated by their robustness values of the probabilistic metric temporal logic (P-MTL) constraints.…”

“…Here, the views are randomly sampled as potential targets via Markov Chain Monte Carlo methods and evaluated by their robustness values of the probabilistic metric temporal logic (P-MTL) constraints. In Section V, we compare our approach against the P-MTL method [3] for the illegal mining search task.…”

“…In our simulated experiment, we use a simple color thresholding technique to detect AoI. For our gold mining experiment, we use a CNN to recognize the mining machines and traces as AoI based on a dataset of remote sensing images taken from Planet Satellite and UAVs as in [3]. 2) Egocentric Map mt: We use previous observations up to time step t to derive the egocentric map mt.…”

“…Developing adaptive exploration algorithms is particularly important in time-critical tasks in outdoor environments such as Search and Rescue (SaR) [2], illegal activity detection (e.g. human-trafficking, gold mining [3]), or environmental applications (localizing wildlife, pollution, or mobile machinery).…”

“…We have tested our approach in simulation on two different settings: 1) 35 randomly simulated maps of AoIs of different shapes and distributions and 2) 19 maps from Planet satellite imagery for detecting illegal gold mining activity in Amazonian forests in Peru [3]. We built a random scenario generator to create the training and testing maps for the first task.…”