Information-Theoretic Planning with Trajectory Optimization for Dense 3D Mapping

Abstract

We propose an information-theoretic planning approach that enables mobile robots to autonomously construct dense 3D maps in a computationally efficient manner. Inspired by prior work, we accomplish this task by formulating an information-theoretic objective function based on Cauchy-Schwarz quadratic mutual information (CSQMI) that guides robots to obtain measurements in uncertain regions of the map. We then contribute a two stage approach for active mapping. First, we generate a candidate set of trajectories using a combination of global planning and generation of local motion primitives. From this set, we choose a trajectory that maximizes the information-theoretic objective. Second, we employ a gradientbased trajectory optimization technique to locally refine the chosen trajectory such that the CSQMI objective is maximized while satisfying the robot's motion constraints. We evaluated our approach through a series of simulations and experiments on a ground robot and an aerial robot mapping unknown 3D environments. Real-world experiments suggest our approach reduces the time to explore an environment by 70% compared to a closest frontier exploration strategy and 57% compared to an information-based strategy that uses global planning, while simulations demonstrate the approach extends to aerial robots with higher-dimensional state. Global plans Local motion primitives (a) Global plans Local motion primitives (b) Global plans Local motion primitives Opt. Global plans Opt. Local motion primitives

Directionally Controlled Time-of-Flight Ranging for Mobile Sensing Platforms

“…Adaptive sampling in 3D models: Adaptive sampling techniques from signal processing [34] are used extensively for efficient mesh representations of computer generated scenes [39,4]. In robotics and vision, information theoretic approaches are used to model adaptive 3D sensing for SLAM and other applications [40,5,15]. In this paper, we are interested in adaptive algorithms for LIDAR sensors that take into account physical constraints such as the power expanded on far away objects or on objects moving out of the field-of-view.…”

“…A unique characteristic of our setup is that we can adapt this motion to the current set of scene measurements. We control the MEMS mirror over the FOV by exploiting strategies used for LIDAR sensing in robotics [18,5,40]. In particular, we first generate a series of candidate trajectories that conform to any desired global physical constraints on the sensor.…”

“…In particular, we first generate a series of candidate trajectories that conform to any desired global physical constraints on the sensor. We then select from these candidates by maximizing a local information theoretic measure that has had success in active vision for robotics [5].…”

“…We use a form for I(m|x t ) in Eq 2 derived from the Cauchy-Schwarz quadratic mutual information (CSQMI) for a single laser beam [5]. The expression for CSQMI is reproduced here from [5],…”

Information-Driven Autonomous Exploration for a Vision-Based Mav

“…Another relevant approach is the one by Charrow et al (2015), which generates a set of trajectories by combining a frontierbased technique and local primitives generated by control sampling, then choose the one that maximizes the information-theoretic objective and further optimize it with a gradient-based optimization. Isler et al (2016), instead, select the next-best-view according to the information gain in a volumetric map.…”