Learning search polices from humans in a partially observable context

Decision making and planning for which the state information is only partially available is a problem faced by all forms of intelligent entities they being either virtual, synthetic or biological. The standard approach to mathematically solve such a decisional problem is to formulate it as a partially observable decision process (POMDP) and apply the same optimisation techniques used in the Markov decision process (MDP). However, applying naively the same methodology to solve MDPs as with POMDPs makes the problem computationally intractable. To address this problem, we take a programming by demonstration approach to provide a solution to the POMDP in continuous state and action space. In this work, we model the decision making process followed by humans when searching blindly for an object on a table. We show that by representing the belief of the human’s position in the environment by a particle filter (PF) and learning a mapping from this belief to their end effector velocities with a Gaussian mixture model (GMM), we can model the human’s search process and reproduce it for any agent. We further categorize the type of behaviours demonstrated by humans as being either risk-prone or risk-averse and find that more than 70% of the human searches were considered to be risk-averse. We contrast the performance of this human-inspired search model with respect to greedy and coastal navigation search methods. Our evaluation metric is the distance taken to reach the goal and how each method minimises the uncertainty. We further analyse the control policy of the coastal navigation and GMM search models and argue that taking into account uncertainty is more efficient with respect to distance travelled to reach the goal.

Learning Search Strategies from Human Demonstrations

Guillaume Pierre Luc De Chambrier

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.

EPFL2016

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

Human and Machine Learning in Non-Markovian Decision Making

Aaron Michael Clarke, Michael Herzog, Silvia Marchesotti, Walter Senn, Elisa Tartaglia

Humans can learn under a wide variety of feedback conditions. Reinforcement learning (RL), where a series of rewarded decisions must be made, is a particularly important type of learning. Computational and behavioral studies of RL have focused mainly on Markovian decision processes, where the next state depends on only the current state and action. Little is known about non-Markovian decision making, where the next state depends on more than the current state and action. Learning is non-Markovian, for example, when there is no unique mapping between actions and feedback. We have produced a model based on spiking neurons that can handle these non-Markovian conditions by performing policy gradient descent [1]. Here, we examine the model's performance and compare it with human learning and a Bayes optimal reference, which provides an upper-bound on performance. We find that in all cases, our population of spiking neurons model well-describes human performance.