Robot Learning From Demonstration

By Christopher G. Atkeson, Stefan Schaal. 1997

How researchers see this paper

The researchers whose trusted networks rank it highest, strongest advocate first. Every researcher gets a different order.

Citations in the corpus

Reinforcement learning in robotics: A survey (2013, positive): “This has resulted in a variety of robot reinforcement learning applications ranging from pendulum swing-up tasks learned with DDP (Atkeson and Schaal, 1997; Atkeson, 1998), devil-sticking (a form of gyroscopic”
Overcoming Exploration in Reinforcement Learning with Demonstrations (2017, positive): “itation Learning: Previous work has combined reinforcement learning with demonstrations. Demonstrations have been used to accelerate learning on classical tasks such as cart-pole swing-up and balance [31]. This work initialized policies and (in model-based methods) initialized forward models with de”
Deep TAMER: Interactive Agent Shaping in High-Dimensional State Spaces (2017, neutral): “odology we propose to study here. One area of research on learning from human interaction that has received a great deal of attention is focused on the speciﬁc problem of learning from demonstration (Schaal 1997; Argall et al. 2009; Hussein et al. 2017). Techniques developed to address this problem ”
SOLAR: Deep Structured Representations for Model-Based Reinforcement Learning (2018, unknown): “beneﬁts compared to model-free RL methods [9, 30, 10]. Many of these prior methods learn global models of the system that are then used for planning, generating synthetic experience, or policy search [3, 33]. These methods require an accurate and reliable model and will typically suffer from modelin”
Neural primitives for motion control (2004, unknown): “5A vivid illustration of this is offered by recent work of Atkeson and Schaal [ 68 ], [69], who studied the task of balancing an inverted pendulum on the hand of a robotic arm.”
Imitating Latent Policies from Observation (2018, unknown): “ng a set of demonstrations. This approach has an extensive breadth of applications and a long history of research, ranging from early successes in autonomous driving [18], to applications in robotics [20, 4] and software agents (e.g. [23, 14]). However, these approaches typically assume that the exp”
Robot learning via socially guided exploration (2007, unknown): “Each drive has a range [0, 1 ], initial value of 0.5, a tendency to drift to 0.0, and a drift magnitude of 0.001 (max change in a time step).”
Reinforcement Learning from Demonstration through Shaping (2015, unknown): “…physical and computational differences between demonstrator and learning agent, and limitations of the demonstrator, typically result in suboptimal demonstrations, compromising the quality of behaviour the agent derives from such demonstrations [Atkeson and Schaal, 1997; Taylor et al., 2011b].”
Observational Learning by Reinforcement Learning (2017, unknown): “e behaviour of other agents [16, 17] and then use this, in conjunction with RL techniques, to optimize this inferred function. While these methods have been successfully applied to a variety of tasks [14, 18, 19, 20, 21, 22, 23, 24, 25], one problematic aspect of both these scenarios is that they al”
Catch & Carry (2019, unknown): “number of demonstrations, it is not sufficient to try to directly mimic the demonstration as there will be some discrepancies when recapitulating the movement which will compound and lead to failure [Atkeson and Schaal 1997; Schaal 1997]. A fairly direct approach involves fitting the demonstrations ”
Movement reproduction and obstacle avoidance with dynamic movement primitives and potential fields (2008, unknown): “Several researchers developed methods for learning from demonstration [ 5 ], [6].”
Flow: Architecture and Benchmarking for Reinforcement Learning in Traffic Control (2017, unknown): “gainst that of a human driver [9], [10]. Similarly, expert demonstrations are often considered the objective in robot learning for a variety of tasks, including locomotion, grasping, and manipulation [11], [12], [13]. In these cases, we may assume knowledge of a good control law and that it is feasi”
Learning from Demonstrations: Is It Worth Estimating a Reward Function? (2013, unknown): “AL can be reduced to classification [7,3,6,11] where the agent tries to mimic the expert policy via a Supervised Learning (SL) method such as classification.”
Real-Time Adversarial Attacks (2019, unknown): “This naturally leads us to use an imitation learning and behavior cloning [Atkeson and Schaal, 1997] strategy to overcome the sparse reward problem.”