Team, Visitors, External Collaborators
Overall Objectives
Research Program
Highlights of the Year
New Software and Platforms
New Results
Bilateral Contracts and Grants with Industry
Partnerships and Cooperations
XML PDF e-pub
PDF e-Pub

Section: New Results

Optimal and Uncertainty-Aware Sensing

Tracking of Rigid Objects of Complex Shapes with a RDB-D Camera

Participants : Agniva Sengupta, Alexandre Krupa, Eric Marchand.

In the context of the iProcess project (see Section 8.3.8), we developed a method for accurately tracking the pose of rigid objects of complex shapes using a RGB-D camera [52]. This method only needs a coarse 3D geometric model of the object of interest represented as a 3D mesh. The tracking of the object is based on a joint minimization of geometric and photometric criteria and more particularly on a combination of point-to-plane distance minimization and photometric error minimization. The concept of successive “keyframes” was also used in this approach for minimizing possible drift of the tracking. The proposed approach was validated on both simulated and real data and the results experimentally demonstrated a better tracking accuracy than existing state-of-the-art 6-DoF object tracking methods, especially when dealing with low-textured objects, multiple coplanar faces, occlusions and partial specularities of the scene.

Deformable Object 3D Tracking based on Depth Information and Coarse Physical Model

Participants : Agniva Sengupta, Alexandre Krupa, Eric Marchand.

This research activity was also carried out in the context of the iProcess project (see Section 8.3.8) and will continue with the recent starting GentleMAN project (see Section 8.3.9). It focusses on the elaboration of approaches able to accurately track in real-time the deformation of soft objects using a RGB-D camera. The state-of-the-art approaches are currently relying on the use of Finite Element Model (FEM) to simulate the physics (mechanical behavior) of the deformable object. However, they suffer from the drawback of being excessively dependent on the accurate knowledge of the physical properties of the object being tracked (Young Modulus, Poisson's ratio, etc). This year, we proposed a first method that only required a coarse physical model of the object based on FEM whose parameters do not need to be precise [53]. The method consists in applying a set of virtual forces on the surface mesh of our coarse FEM model in such a way that it deforms to fit the current shape of the object. A point-to-plane distance error between the point cloud provided by the depth camera and the model mesh is iteratively minimized with respect to these virtual forces. The point of application of force is determined by an analysis of the error obtained from rigid tracking, which is done in parallel with the non-rigid tracking. The approach has been validated on simulated objects with ground-truth, as well on real objects of unknown physical properties and experimentally demonstrated that accurate tracking of deformable objects can be achieved without the need of a precise physical model.

Trajectory Generation for Optimal State Estimation

Participants : Marco Cognetti, Paolo Robuffo Giordano.

This activity addresses the general problem of active sensing where the goal is to analyze and synthesize optimal trajectories for a robotic system that can maximize the amount of information gathered by the (few) noisy outputs (i.e., sensor readings) while at the same time reducing the negative effects of the process/actuation noise. Over the last years we have developed a general framework for solving online the active sensing problem by continuously replanning an optimal trajectory that maximizes a suitable norm of the Constructibility Gramian (CG), while also coping with a number of constraints including limited energy and feasibility. The results obtained so far have been generalized and summarized in [27], where the online trajectory replanning for CG maximization has been applied to two relevant case studies (unicycle and quadrotor) and validated via a large statistical campaign. We are actually working towards the extension of this machinery to the case of realization of a robot task (e.g., reaching and grasping for a mobile manipulator), and to the mutual localization problem for a multi-robot group.

Robotic manipulators in Physical Interaction with the Environment

Participant : Claudio Pacchierotti.

As robotic systems become more flexible and intelligent, they must be able to move into environments with a high degree of uncertainty or clutter, such as our homes, workplaces, and the outdoors. In these unstructured scenarios, it is possible that the body of the robot collides with its surroundings. As such, it would be desirable to characterise these contacts in terms of their location and interaction forces. We worked to address the problem of detecting and isolating collisions between a robotic manipulator and its environment, using only on-board joint torque and position sensing [37]. We presented an algorithm based on a particle filter that, under some assumptions, is able to identify the contact location anywhere on the robot body. It requires the robot to perform small exploratory movements, progressively integrating the new sensing information through a Bayesian framework. The method assumes negligible friction forces, convex contact surfaces, and linear contact stiffness. Compared to existing approaches, it allows this detection to be carried in almost all the surface of the robot's body. We tested the proposed approach both in simulation and in a real environment. Experiments in simulation showed that our approach outperformed two other methods that made simpler assumptions. Experiments in a real environment using a robot with joint torque sensors showed the applicability of the method to real world scenarios and its ability to cope with situations where the algorithm's assumptions did not hold.

Cooperative Localization using Interval Analysis

Participants : Ide Flore Kenmogne Fokam, Vincent Drevelle, Eric Marchand.

In the context of multi-robot fleets, cooperative localization consists in gaining better position estimate through measurements and data exchange with neighboring robots. Positioning integrity (i.e., providing reliable position uncertainty information) is also a key point for mission-critical tasks, like collision avoidance. The goal of this work is to compute position uncertainty volumes for each robot of the fleet, using a decentralized method (i.e., using only local communication with the neighbors). The problem is addressed in a bounded-error framework, with interval analysis and constraint propagation methods. These methods enable to provide guaranteed position error bounds, assuming bounded-error measurements. They are not affected by over-convergence due to data incest, which makes them a well sound framework for decentralized estimation. Quantifier elimination techniques have been used to consider uncertainty in the landmarks positions without adding pessimism in the computed solution. This work has been applied to cooperative localization of UAVs, based on image and range measurements [20].