doc. Ing. Karel Zimmermann, Ph.D.

Accurate perception, which enables explainable reasoning about possible outcomes of the robot-environment interaction, is essential for many fundamental capabilities, such as localization, mapping, planning or control. Despite several successful robotic's solutions, a robust perception architecture that would allow for efficient self-supervised adaptation in an unknown environment remains an open research problem. We claim that the main issues that prevent one from building such architectures are the following: (i) Black-box architectures that neglect the physical embodiment of the robot, such as deep NNs, suffer from poor generalization, weak explainability, catastrophic forgetting and the impossibility of being transferred among different robotics platforms and environments; (ii) Consequent sample-inefficient learning requires a huge number of expensive, domain-specific, human-annotated data that cannot be transferred to other domains; (iii) Resulting complex architectures have a vast number of parameters that have to be adapted/learned jointly to avoid undesirable interference among independently tuned components. We aim to address these issues by designing grey-box, explainable, embodiment-aware, and end-to-end differentiable architectures that enable joint self-supervised adaptation from any successively incoming onboard measurements in a probabilistically consistent way.

Accurate local 3D perception is an essential component for many fundamental capabilities such as emergency braking, active damping or self-localization from offline maps. Consequently, all autonomous vehicles require a sensor providing high resolution and long range 3D measurements. Since state-of-the-art rotating lidars are very expensive, heavy and contain moving parts prone to mechanical wear, several manufacturers have announced development of cheaper, lighter, smaller and motionless Solid State Lidars (SSL). SSLs should become available before the end of 2017 with target cost of $250 at automotive scale production, which make them affordable for ordinary cars in a close future. In contrast to rotating lidars, the SSL can independently steer pulses of light by shifting and focus its attention on the parts of the scene important for the current task. Task-driven reactive control of hundreds of thousands rays per second using only an on-board computer is a challenging problem, which calls for highly efficient parallelizable algorithms. We are looking for students who want to cooperate with us in the research of active mapping/segmentation/detection methods, which simultaneously (i) learns to reconstruct a dense 3D map from sparse depth measurements and (ii) optimize the reactive control of depth-measuring rays in order to minimize reconstruction error. Další informace lze nalézt na http://cmp.felk.cvut.cz/~zimmerk/