Spatio-temporal exploration strategies for long-term autonomy of mobile robots

Abstract

We present a study of spatio-temporal environment representations and exploration strategies for long-term deployment of mobile robots in real-world, dynamic environments.\ud We propose a new concept for life-long mobile robot spatio-temporal exploration that aims at building, updating and maintaining the environment model during the long-term deployment.\ud The addition of the temporal dimension to the explored space makes the exploration task a never-ending data-gathering process, which we address by application of information-theoretic exploration techniques to world representations that model the uncertainty of environment states as probabilistic functions of time.\ud We evaluate the performance of different exploration strategies and temporal models on real-world data gathered over the course of several months.\ud The combination of dynamic environment representations with information-gain exploration principles allows to create and maintain up-to-date models of continuously changing environments, enabling efficient and self-improving long-term operation of mobile robots

Go with the Flow: Exploration and Mapping of Pedestrian Flow Patterns from Partial Observations

“…In [9] and [10], for example, the decision making for environmental surveillance and monitoring is based on Gaussian Processes, which allow the robot to learn the temporal patterns in the environment. Other approaches [11], [12] are instead based on the assumption that some of the environmental variations observed are caused by people's daily routines; [11] present a method for life-long spatiotemporal exploration of dynamic environments, using the entropy of binary state predictions in an occupancy map to create a scheduler that determines which areas and times to explore for each day. However, reasoning about human motion requires more complex model representations, and binary states are not sufficient to describe the pedestrian flows covered in this paper.…”

Learning Autonomous Exploration and Mapping with Semantic Vision

“…There is a large body of literature on autonomous exploration and mapping [13,27,32]. Typically, the existing pipelines consist of three stages: identifying the unexplored regions, determining which region to go next and navigating to that region.…”

Curiosity Did Not Kill the Robot

“…Kaplan and Oudeyer demonstrated that a robot is able to explore within its repertoire of motor primitives (Kaplan and Oudeyer 2011). Action selection strategies based on information gain have also been applied for spatiotemporal exploration by mobile robots in continuously changing environments (Müller et al 2014;Santos et al 2017). The interactive art sculpture presented in Chan et al (2015) is a distributed system with a large sensorimotor space that employed a similar concept of curiositybased learning.…”