Learning Search Strategies from Human Demonstrations

Decision making and planning with partial state information is a problem faced by all forms of intelligent entities. The formulation of a problem under partial state information leads to an exorbitant set of choices with associated probabilistic outcomes making its resolution difficult when using traditional planning methods. Human beings have acquired the ability of acting under uncertainty through education and self-learning. Transferring our know-how to artificial agents and robots will make it faster for them to learn and even improve upon us in tasks in which incomplete knowledge is available, which is the objective of this thesis. We model how humans reason with respect to their beliefs and transfer this knowledge in the form of a parameterised policy, following a Programming by Demonstration framework, to a robot apprentice for two spatial navigation tasks: the first task consists of localising a wooden block on a table and for the second task a power socket must be found and connected. In both tasks the human teacher and robot apprentice only rely on haptic and tactile information. We model the human and robot's beliefs by a probability density function which we update through recursive Bayesian state space estimation. To model the reasoning processes of human subjects performing the search tasks we learn a generative joint distribution over beliefs and actions (end-effector velocities) which were recorded during the executions of the task. For the first search task the direct mapping from belief to actions is learned whilst for the second task we incorporate a cost function used to adapt the policy parameters in a Reinforcement Learning framework and show a considerable improvement over solely learning the behaviour with respect to the distance taken to accomplish the task. Both search tasks above can be considered as active localisation as the uncertainty originates only from the position of the agent in the world. We consider searches in which both the position of the robot and features of the environment are uncertain. Given the unstructured nature of the belief a histogram parametrisation of the joint distribution of the robots position and features is necessary. However, naively doing so becomes quickly intractable as the space and time complexity is exponential. We demonstrate that by only parametrising the marginals and by memorising the parameters of the measurement likelihood functions we can recover the exact same solution as the naive parametrisations at a cost which is linear in space and time complexity.

From High-Level to Low-Level Robot Learning of Complex Tasks: Leveraging Priors, Metrics and Dynamical Systems

Nadia Barbara Figueroa Fernandez

Humans have a remarkable way of learning, adapting and mastering new manipulation tasks. With the current advances in Machine Learning (ML), the promise of having robots with such capabilities seems to be on the cusp of reality. Transferring human-level skills to robots, however, is complicated as they involve a level of complexity that cannot be tackled by classical ML methods in an unsupervised way. Such complexities involve: (i) automatically decomposing tasks into control-oriented encodings, (ii) extracting invariances and handling idiosyncrasies of data acquired from human demonstrations, and (iii) learning models that guarantee stability and convergence. In this thesis, we push the boundaries in the learning from demonstration (LfD) domain by addressing these challenges with minimal human intervention and parameter tuning. We introduce novel learning approaches based on Bayesian non-parametrics and kernel-methods while encoding novel metrics and priors, leveraging them with dynamical systems (DS) theory. In the first part of this thesis we focus on learning complex sequential manipulation tasks from heterogeneous and unstructured demonstrations. Such tasks are those composed of a sequence of discrete actions. The particular challenge is learning these tasks without any prior knowledge on the number of actions (unstructuredness) or restrictions as to how the human is demonstrating the task (heterogeneous), e.g. changes in reference frames. We propose a Bayesian non-parametric learning framework that can jointly segment and discover unique discrete actions (and their sequence) in continuous demonstrations of the task. Hence, we learn an entire task from a continuous, unrestricted natural demonstration. The learned structure of the complex tasks and the segmented data were then used to parametrize a hybrid controller to execute two cooking tasks of increasing complexity: (i) single-arm pizza dough rolling, and (ii) a dual-arm vegetable peeling task. Throughout this thesis, we assume that both the human and robot motions are driven by autonomous state-dependent DS. Hence, in the second part of this thesis we offer two novel DS formulations to represent and execute a complex task. We begin by proposing a DS-based motion generator formulation and learning scheme that is capable of automatically encoding continuous and complex motions while ensuring global asymptotic stability. The type of tasks that can be learned with this approach are unparalleled to previous work in DS-based LfD and are validated on production line and household activities. Further, we propose a novel DS formulation and learning scheme that can encode both complex motions and varying impedance requirements along the task; i.e., the robot must be compliant in some regions of the task, while stiff in others. This approach is validated on trajectory tracking tasks where a robot arm must precisely draw letters on a surface. Due to the generalization power and straight-forward learning schemes of the proposed DS-based motion generators, we also apply them to more complex application. Such applications include providing adaptive (i) navigation strategies for mobile agents, and (ii) locomotion and co-manipulation tasks of biped robots. These applications are particularly novel in the LfD domain, as most work is solely focused on robotic arm control. In the last part of the thesis, we explore learning complex behaviors in joint space for single and multi-arm systems.

EPFL2019

Using reinforcement learning for optimizing the reproduction of tasks in robot programming by demonstration

Florent Günter

As robots start pervading human environments, the need for new interfaces that would simplify human-robot interaction has become more pressing. Robot Programming by Demonstration (RbD) develops intuitive ways of programming robots, taking inspiration in strategies used by humans to transmit knowledge to apprentices. The user-friendliness of RbD is meant to allow lay users with no prior knowledge in computer science, electronics or mechanics to train robots to accomplish tasks the same way as they would with a co-worker. When a trainer teaches a task to a robot, he/she shows a particular way of fulfilling the task. For a robot to be able to learn from observing the trainer, it must be able to learn what the task entails (i.e. answer the so-called "What-to-imitate?" question), by inferring the user's intentions. But most importantly, the robot must be able to adapt its own controller to fit at best the demonstration (the so-called "How-to-imitate?" question) despite different setups and embodiments. The latter is the question that interested us in this thesis. It relates to the problem of optimizing the reproduction of the task under environmental constraints. The "How-to-imitate?" question is subdivided into two problems. The first problem, also known as the "correspondence problem", relates to resolving the discrepancy between the human demonstrator and robot's body that prevent the robot from doing an identical reproduction of the task. Even though we helped ourselves by considering solely humanoid platforms, that is platforms that have a joint configuration similar to that of the human, discrepancies in the number of degrees of freedom and range of motion remained. We resolved these by exploiting the redundant information conveyed through the demonstrations by collecting data through different frames of reference. By exploiting these redundancies in an algorithm comparable to the damped least square algorithm, we are able to reproduce a trajectory that minimizes the error between the desired trajectory and the reproduced trajectory across each frame of reference. The second problem consists in reproducing a trajectory in an unknown setup while respecting the task constraints learned during training. When the information learned from the demonstration no longer suffice to generalize the task constraints to a new set-up, the robot must re-learn the task; this time through trial-and-error. Here we considered the combination of trial-and-error learning to complement RbD. By adding a trial-and-error module to the original Imitation Learning algorithm, the robot can find a solution that is more adapted to the context and to its embodiment than the solution found using RbD. Specifically, we compared Reinforcement Learning (RL) – to other classical optimization techniques. We show that the system is advantageous in that: a) learning is more robust to unexpected events that have not been encountered during the demonstrations and b) the robot is able to optimize its own model of the task according to its own embodiment.

EPFL2009

Using reinforcement learning for optimizing the reproduction of tasks in robot programming by demonstration

Florent Günter

EPFL2009

From High-Level to Low-Level Robot Learning of Complex Tasks: Leveraging Priors, Metrics and Dynamical Systems

Nadia Barbara Figueroa Fernandez

EPFL2019