An Object Template Approach to Manipulation for Semi-autonomous Avatar Robots

Nowadays, the first steps towards the use of mobile robots to perform manipulation tasks in remote environments have been made possible. This opens new possibilities for research and development, since robots can help humans to perform tasks in many scenarios. A remote robot can be used as avatar in applications such as for medical or industrial use, in rescue and disaster recovery tasks which might be hazardous environments for human beings to enter, as well as for more distant scenarios like planetary explorations. Among the most typical applications in recent years, research towards the deployment of robots to mitigate disaster scenarios has been of great interest in the robotics field. Disaster scenarios present challenges that need to be tackled. Their unstructured nature makes them difficult to predict and even though some assumptions can be made for human-designed scenarios, there is no certainty on the expected conditions. Communications with a robot inside these scenarios might also be challenged; wired communications limit reachability and wireless communications are limited by bandwidth. Despite the great progress in the robotics research field, these difficulties have prevented the current autonomous robotic approaches to perform efficiently in unstructured remote scenarios. On one side, acquiring physical and abstract information from unknown objects in a full autonomous way in uncontrolled environmental conditions is still an unsolved problem. Several challenges have to be overcome such as object recognition, grasp planning, manipulation, and mission planning among others. On the other side, purely teleoperated robots require a reliable communication link robust to reachability, bandwidth, and latency which can provide all the necessary feedback that a human operator needs in order to achieve sufficiently good situational awareness, e.g., worldmodel, robot state, forces, and torques exerted. Processing this amount of information plus the necessary training to perform joint motions with the robot represent a high mental workload for the operator which results in very low execution times. Additionally, a pure teleoperated approach is error-prone given that the success in a manipulation task strongly depends on the ability and expertise of the human operating the robot. Both, autonomous and teleoperated robotic approaches have pros and cons, for this reason a middle ground approach has emerged. In an approach where a human supervises a semi-autonomous remote robot, strengths from both, full autonomous and purely teleoperated approaches can be combined while at the same time their weaknesses can be tackled. A remote manipulation task can be divided into sub-tasks such as planning, perception, action, and evaluation. A proper distribution of these sub-tasks between the human operator and the remote robot can increase the efficiency and potential of success in a manipulation task. On the one hand, a human operator can trivially plan a task (planning), identify objects in the sensor data acquired by the robot (perception), and verify the completion of a task (evaluation). On the other hand, it is challenging to remotely control in joint space a robotic system like a humanoid robot that can easily have over 25 degrees of freedom (DOF). For this reason, in this approach the complex sub-tasks such as motion planning, motion execution, and obstacle avoidance (action) are performed autonomously by the remote robot. With this distribution of tasks, the challenge of converting the operator intent into a robot action arises. This thesis investigates concepts of how to efficiently provide a remote robot with the operator intent in a flexible means of interaction. While current approaches focus on an object-grasp-centered means of interaction, this thesis aims at providing physical and abstract properties of the objects of interest. With this information, the robot can perform autonomous subtasks like locomotion through the environment, grasping objects, and manipulating them at an affordance-level avoiding collisions with the environment in order to efficiently accomplish the manipulation task needed. For this purpose, the concept of Object Template (OT) has been developed in this thesis. An OT is a virtual representation of an object of interest that contains information that a remote robot can use to manipulate such object or other similar objects. The object template concept presented here goes beyond state-of-the-art related concepts by extending the robot capabilities to use affordance information of the object. This concept includes physical information (mass, center of mass, inertia tensor) as well as abstract information (potential grasps, affordances, and usabilities). Because humans are very good at analysing a situation, planning new ways of how to solve a task, even using objects for different purposes, it is important to allow communicating the planning and perception performed by the operator such that the robot can execute the action based on the information contained in the OT. This leverages human intelligence with robot capabilities. For example, as an implementation in a 3D environment, an OT can be visualized as a 3D geometry mesh that simulates an object of interest. A human operator can manipulate the OT and move it so that it overlaps with the visualized sensor data of the real object. Information of the object template type and its pose can be compressed and sent using low bandwidth communication. Then, the remote robot can use the information of the OT to approach, grasp, and manipulate the real object. The use of remote humanoid robots as avatars is expected to be intuitive to operators (or potential human response forces) since the kinematic chains and degrees of freedom are similar to humans. This allows operators to visualize themselves in the remote environment and think how to solve a task, however, task requirements such as special tools might not be found. For this reason, a flexible means of interaction that can account for allowing improvisation from the operator is also needed. In this approach, improvisation is described as "a change of a plan on how to achieve a certain task, depending on the current situation". A human operator can then improvise by adapting the affordances of known objects into new unknown objects. For example, by utilizing the affordances defined in an OT on a new object that has similar physical properties or which manipulation skills belong to the same class. The experimental results presented in this thesis validate the proposed approach by demonstrating the successful achievement of several manipulation tasks using object templates. Systematic laboratory experimentation has been performed to evaluate the individual aspects of this approach. The performance of the approach has been tested in three different humanoid robotic systems (one of these robots belongs to another research laboratory). These three robotic platforms also participated in the renowned international competition DARPA Robotics Challenge (DRC) which between 2012 and 2015 was considered the most ambitious and challenging robotic competition