Towards Autonomous Planetary Exploration Collaborative Multi-Robot Localization and Mapping in GPS-denied Environments

Transcription

1 DLR.de Chart 1 International Technical Symposium on Navigation and Timing (ITSNT) Toulouse, France, 2017 Towards Autonomous Planetary Exploration Collaborative Multi-Robot Localization and Mapping in GPS-denied Environments Martin J. Schuster Department of Perception and Cognition Robotics and Mechatronics Center DLR German Aerospace Center

2 DLR.de Chart 2 Scenarios for Heterogeneous Multi-Robot Teams In Space: Planetary Exploration On Earth: Search & Rescue Missions

3 DLR.de Chart 3 The Lightweight Rover Unit (LRU) for Planetary Exploration [M. J. Schuster et al., Towards Autonomous Planetary Exploration: The Lightweight Rover Unit (LRU), its Success in the SpaceBotCamp Challenge, and Beyond, in Journal of Intelligent & Robotic Systems (JINT), 2017] [A. Wedler et al., First Results of the ROBEX Analogue Mission Campaign: Robotic Deployment of Seismic Networks for Future Lunar Missions, in International Astronautical Congress (IAC), 2017]

4 DLR.de Chart 4 Heterogeneous Multi-Robot Teams for Autonomous Exploration

5 DLR.de Chart 5 Mapping with two Rovers

6 DLR.de Chart 6 Navigation for Multi-Robot Teams in Previously Unknown Environments 1. Real-time Local State Estimation & Obstacle Avoidance 2. Online Global Localization and Mapping (6D-SLAM) Local Reference Filter Terrain Classification Exchange of Submaps Multi-Robot Graph How to combine these Methods? Novel graph topology for decoupled integration of filter estimates Intra- and inter-robot loop closures by matching submaps Integration into modular software architecture

7 DLR.de Chart 7 Architecture

8 DLR.de Chart 8 Local Reference Filter Spinoff: Roboception GmbH Extended Kalman Filter (EKF) for real-time state estimation System state: x = p i T, v i T, q i T, b a T, b ω T R 16 Error state: δ = (δp i T, δv i T, δσ i T, δb a T, δ ω T ) R 15 Time-delay compensation for visual odometry (VO) measurements Locally drift-free estimation through VO keyframes Switch frame of reference into current robot pose Reset pose uncertainty estimates to zero Transform other filter state variables into new frame of reference Benefits of frame switches: Reset unbounded covariances for unobservable states Long-term numerical stability No violation of small-angle approximations for error-state filter Improved integration of filter estimates into SLAM graph [K. Schmid et al., Local Reference Filter for Life-Long Vision Aided Inertial Navigation, FUSION, 2014]

9 DLR.de Chart 9 3D Submap Creation and Matching Creation of Submaps Local aggregation of dense 3D data Distributed computation Low bandwidth requirements for exchange Submap Matching Estimation of transformation between submaps Initial Alignment: Keypoint selection: Obstacle classification Feature matching: CSHOT (local geometry and texture) Refinement: Iterative Closest Point (ICP) Inter- and Intra-Robot Loop Closures [C. Brand et al., Submap Matching for Stereo-Vision Based Indoor/Outdoor SLAM, IROS, 2015]

10 DLR.de Chart 10 Multi-Robot Graph Topology for Sequential Odometry Measurements [M. J. Schuster et al., Multi-Robot 6D Graph SLAM Connecting Decoupled Local Reference Filters, IROS, 2015]

11 DLR.de Chart 11 Multi-Robot Graph Topology for Sequential Odometry Measurements [M. J. Schuster et al., Multi-Robot 6D Graph SLAM Connecting Decoupled Local Reference Filters, IROS, 2015]

12 DLR.de Chart 12 Multi-Robot Graph Topology for Sequential Odometry Measurements [M. J. Schuster et al., Multi-Robot 6D Graph SLAM Connecting Decoupled Local Reference Filters, IROS, 2015]

13 DLR.de Chart 13 New Multi-Robot Graph Topology for Local Reference Filter Estimates [M. J. Schuster et al., Multi-Robot 6D Graph SLAM Connecting Decoupled Local Reference Filters, IROS, 2015]

14 DLR.de Chart 14 Incremental Multi-Robot Graph SLAM SLAM Front-End (Data Association) Loosely coupled integration of filter estimates according to their independence assumptions No knowledge about filter-internal states required Aggregation of high-frequency measurements Small graph: Fast optimization steps SLAM Back-End (Graph Optimization) Incremental non-linear optimization (lest-squares) f Θ = i f i (Θ i ) arg min Θ log f Θ = arg min Θ 1 2 i h i Θ i 2 z i Σi Robust cost function for outlier suppression (Cauchy)

15 DLR.de Chart 15 Mapping with two Rovers Area: 57m x 53m SLAM-Graph: 105 Nodes, 143 Factors LRU1 LRU2 Left Camera Left Camera Depth Image Depth Image

16 DLR.de Chart 16 ICARSC 2016 Best Paper Award SpaceBotCamp 2015 National Robotics Challenge Moon-like Scenario Area: 13m x 18m [M. J. Schuster et al., Towards Autonomous Planetary Exploration: The Lightweight Rover Unit (LRU), its Success in the SpaceBotCamp Challenge, and Beyond, in Journal of Intelligent & Robotic Systems (JINT), 2017]

17 DLR.de Chart 17 Autonomous Exploration Goal selection according to expected information gain Active loop closing: Trade-off between exploration of unknown areas and exploitation of known information Consideration of match effect, impact of a loop closure, as well as match likelihood and match costs [H. Lehner et al., Exploration with Active Loop Closing: A Trade-off between Exploration Efficiency and Map Quality, in IROS, 2017]

18 DLR.de Chart 18 Autonomous Exploration

19 DLR.de Chart 19 Autonomous Exploration

20 DLR.de Chart 20 Long-Range Navigation & Multi-Robot Exploration Experiments ROBEX Project ( ) Moon-analogue environment on Mt. Etna [A. Wedler et al., First Results of the ROBEX Analogue Mission Campaign: Robotic Deployment of Seismic Networks for Future Lunar Missions, in International Astronautical Congress (IAC), 2017]

21 DLR.de Chart 21 ROBEX: Multi-Robot Exploration Experiment on Mt. Etna

22 DLR.de Chart 22 Future Work and Outlook ARCHES Project ( ) Heterogeneous teams of robots Long-term and large-scale localization and mapping

23 DLR.de Chart 23 References [1] A. Wedler, B. Rebele, J. Reill, M. Suppa, H. Hirschmüller, C. Brand, M. Schuster, B. Vodermayer, H. Gmeiner, A. Maier, B. Willberg, K. Bussmann, F. Wappler, M. Hellerer and R. Lichtenheldt, LRU - Lightweight Rover Unit, in Symposium on Advanced Space Technologies in Robotics and Automation (ASTRA) (Noordwijk, Netherlands, 2015). [2] M. J. Schuster, C. Brand, S. G. Brunner, P. Lehner, J. Reill, S. Riedel, T. Bodenmüller, K. Bussmann, S. Büttner, A. Dömel, W. Friedl, I. Grixa, M. Hellerer, H. Hirschmüller, M. Kassecker, Z.-C. Márton, C. Nissler, F. Ruess, M. Suppa and A. Wedler, The LRU Rover for Autonomous Planetary Exploration and its Success in the SpaceBotCamp Challenge, in IEEE International Conference on Autonomous Robot Systems and Competitions (ICARSC) (2016). [3] M. J. Schuster, S. G. Brunner, K. Bussmann, S. Büttner, A. Dömel, M. Hellerer, H. Lehner, P. Lehner, O. Porges, J. Reill, S. Riedel, M. Vayugundla, B. Vodermayer, T. Bodenmüller, C. Brand, W. Friedl, I. Grixa, H. Hirschmüller, M. Kaßecker, Z.-C. Márton, C. Nissler, F. Ruess, M. Suppa and A. Wedler, Towards Autonomous Planetary Exploration: The Lightweight Rover Unit (LRU), its Success in the SpaceBotCamp Challenge, and Beyond, Journal of Intelligent & Robotic Systems (JINT) (2017). [4] M. J. Schuster, C. Brand, H. Hirschmüller, M. Suppa and M. Beetz, Multi-Robot 6D Graph SLAM Connecting Decoupled Local Reference Filters, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (Hamburg, Germany, 2015). [5] K. Schmid, T. Tomić, F. Ruess, H. Hirschmüller and M. Suppa, Stereo Vision based indoor / outdoor Navigation for Flying Robots, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (Tokyo, Japan, 2013), pp [6] K. Schmid, P. Lutz, T. Tomić, E. Mair and H. Hirschmüller, Autonomous Vision-based Micro Air Vehicle for Indoor and Outdoor Navigation, Journal of Field Robotics 31 (2014) [7] K. Schmid, F. Ruess and D. Burschka, Local Reference Filter for Life-Long Vision Aided Inertial Navigation, International Conference on Information Fusion (2014). [8] C. Brand, M. J. Schuster, H. Hirschmüller and M. Suppa, Submap Matching for Stereo-Vision Based Indoor/Outdoor SLAM, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (Hamburg, Germany, 2015). [9] H. Lehner, M. J. Schuster, T. Bodenmüller and S. Kriegel, Exploration with Active Loop Closing: A Trade-off between Exploration Efficiency and Map Quality, in IEEE/RSJ International Conference on Intelligent Robots and Systems (IROS) (2017). [10] A. Wedler, M. Vayugundla, H. Lehner, P. Lehner, M. Schuster, S. Brunner, W. Stürzl, A. Dömel, H. Gmeiner, B. Vodermayer, B. Rebele, I. Grixa, K. Bussmann, J. Reill, B. Willberg, A. Maier, P. Meusel, F. Steidle, M. Smisek, M. Hellerer, M. Knapmeyer, F. Sohl, A. Heffels, L. Witte, C. Lange, R. Rosta, N. Toth, S. Völk, A. Kimpe, P. Kyr, et al., First Results of the ROBEX Analogue Mission Campaign: Robotic Deployment of Seismic Networks for Future Lunar Missions, in International Astronautical Congress (IAC) (2017).