Challenges in 3D motion planning and control for articulated tracked robots in Urban Search & Rescue
Arun Kumar Singh, Mario Gianni, Federico Ferri
Mercoledì, 3 Giugno, 2015 - 10:00
Aula B203, Via Ariosto 25
Articulated tracked robots are currently used in search and rescue, military, agricultural
and planetary exploration applications, where terrain conditions are difficult and unpredictable. They are better suited for such tasks than wheeled vehicles due to the larger contact area of tracks with the ground, which provides better traction on harsh terrains. As such robots are deployed in dynamic, unstructured, and open environments, such as a disaster scenario, a key requirement is the ability to interpet the environment, plan highly efficient motions and execute these motions for autonomous safe navigation and morphological adaptation. Several research efforts in robotics have been made to increase the level of autonomy of articulated tracked vehicles, focusing on adaptation, stability, self-reconfiguration, track-soil interaction, motion planning and control.
The aim of this workshop is to bring together researchers of different areas and students interested in the subject to discuss the main open research challenges in 3D motion planning and control of these robots in Urban Search & Rescue.
The workshop includes talks from three speakers that are active in the different subareas of this topic and that will present their research work as well as their in-field experience in deploying these robots in real disaster scenarios.
First talk: 10:00 - 10:45
Speaker: Dr. Arun Kumar Singh, Post-doctoral researcher, Bio-Medical Robotics Lab, BGU, Israel.
Title: Some fresh ideas for motion planning on uneven/rough terrains.
Abstract: In this talk, I will discuss some key challenges associated with motion planning on uneven terrains. In particular I will show, how the complexity is decided by how rigorously we model the robot-terrain interaction. If we model interaction in terms of only postural changes induced on the robot by the underlying terrain, the planning task remains fairly simple. However, sometimes it becomes necessary to incorporate a more complex dynamic robot-terrain interaction during motion planning and this leads to a significant increase in the complexity of the problem. I will show that the first key step in performing a dynamic motion planning on uneven terrain is to decipher how for a particular control input, the robot-terrain interacts to lead to a particular state evolution. Thereafter, I will show that once we have figured out this state evolution puzzle, the task of navigating through an uneven/rough terrain can be framed as an optimization with state and control dependent differential constraints. The peculiar and challenging part is that these constraints rarely have an analytical form, thus, forcing us to adopt an sampling approach to decide whether a particular state and control combination satisfy the differential constraints. To this end, I will present a concept called Feasible Acceleration Count which can be exploited to develop a greedy sampling based motion planner. I will also talk about a new concept called non-linear time scaling which can be exploited to efficiently solve differential constraints and thus finds strong application in uneven terrain motion planning. Towards the end of the talk, I will discuss some current open questions and promising future research directions.
Short bio: Dr.Arun Kumar Singh obtained his Bachelors in Mechanical Engineering from National Institute of Technology, Durgapur, India and his PhD in Electronics and Communication Engineering from Robotics Research Center, IIIT-Hyderabad, India. Currently he is a Postdoctoral fellow at the ABC Robotics Initiative of Ben-Gurion University, Israel. His research interests lies in motion planning and control of complex dynamical systems. At present he is trying to develop human-in-loop motion planning methodologies for reducing the cognitive load of the operator in tele-operation tasks, with special focus towards Robot Assisted Surgeries. He is also highly interested in fusing conventional machine learning algorithms and findings of neuroscience communities for human motion prediction in complex scenarios
Second talk: 11:00 - 11:45
Speaker: Prof. Mario Gianni, DIAG "A. Ruberti", Sapienza University of Rome, Italy.
Title: Adaptive robust three-dimensional trajectory tracking for actively articulated tracked vehicles.
Abstract: In this talk, I will present a unified framework for trajectory tracking controller design of Actively Articulated Tracked Vehicles. The approach develops both a direct and differential kinematic model of the AATV correlating the robot body motion to the sub-tracks motion. The benefit of this approach is to allow the controller to flexibly manage all the DOF of the AATV as well as the steering. The differential kinematic model integrates a differential drive robot model, compensating the slippage between the vehicle tracks and the traversed terrain. The underlying feedback control law dynamically accounts for the kinematic singularities of the mechanical vehicle structure. The designed controller also integrates a strategy selector to reduce both the effort of the sub-tracks
servo motors and the traction force on the robot body, recognizing when the robot is moving on horizontal plane surfaces. According to this strategy, rotational motions of the robot, moving within narrow passages, are also facilitated. This framework has been used for the design of the trajectory tracking controller of the actively articulated racked vehicle Absolem, recently realized for the EU-FP7-ICT Project NIFTi. Several experiments have been performed, in both virtual and real scenarios, to validate the designed trajectory tracking controller, while the robot Absolem negotiates rubbles, stairs and complex terrain surfaces.
Short bio: Mario Gianni is Assistant Professor at the Department of Computer Control, and Management Engineering “A. Ruberti”, Sapienza University of Rome. He is a member of the research group at the Vision, Perception and Learning Robotics Laboratory, ALCOR, directed by Prof. Fiora Pirri. From January 2010 till December 2013 he has been working for the EU FP7 ICT 247870 project NIFTI, aiming at providing natural cooperation between humans and teams of robots in Urban Search and Rescue scenarios. Currently, he is working for the EU FP7 ICT 609763 project TRADR, aiming at developing new S&T for long-term human-robot teaming for robot assisted disaster response. In 2012 he tightly collaborated with the Italian National Fire Corps in order to deploy a human-robot team for assessing damage of historical buildings and cultural artifacts of the ancient city of Mirandola, in Northern Italy, hit by an earthquake. He received the Ph.D. in Computer Engineering from Sapienza University of Rome, with a dissertation titled “Multilayered Cognitive Control for Unmanned Ground Vehicles”. Research interests include statistics and logic, applied to robotics, autonomous navigation and adaptation for self-reconfigurable robots in cluttered environments, low and high-level control in multi-robot collaboration.
Third talk: 12:00 - 12:45
Speaker: Federico Ferri, PhD student, DIAG "A. Ruberti", Sapienza University of Rome, Italy.
Title: 3D motion planning with dynamics
Abstract: In motion planning with dynamics, the objective is to compute a trajectory to the goal region that not only avoids collisions with obstacles but also satisfies differential constraints imposed by robot dynamics. It is motivated by navigation, exploration, search and rescue missions, where it is essential to compute trajectories that can be followed by the robot in a physical real world. Motion planning, in its early years, did not take dynamics into account. Instead, it considered only the geometry of the robot and of obstacles. This simplified view fueled research in sampling-based approaches, such as probabilistic roadmap, rapidly-exploring random tree and expansive space tree methods. However, it has been noted that when dealing with challenges in problems with dynamics, sampling-based motion planners slow down significantly. In this talk, I will describe an approach to incorporates dynamics into sampling-based methods for real-time motion planning for articulated tracked robots. This approach makes use of a motion planning algorithm for sampling primitives in the control space of the robot and relies on a real-time physics engine which performs a state propagation process to estimates the resulting robot state. I'll illustrate the framework which synergically combines motion planning and the underlying phisical simulation, aslo discussing pro and cons.
Short bio: Federico Ferri received his M.Sc. in Engineering in Artificial Intelligence and Robotics from the Department of Computer Control, and Management Engineering “A. Ruberti”, Sapienza University of Rome. He is a member of the research group at the Vision, Perception and Learning Robotics Laboratory, ALCOR, directed by Prof. Fiora Pirri. Currently, he is a PhD student at Sapienza University of Rome. He is working for the H2020-ICT Project SecondHands aiming at designing a robot assistant for maintenance tasks that either pro-actively or as a result of prompting, can offer assistance to maintenance technicians performing routine and preventative maintenance. Research interests include reasoning, knowledge representation and robot planning.