Robot shaping: developing situated agents through learning

By M. Colombetti, Marco Dorigo. 1992

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, background): “Other examples include foraging (Matarić, 1997) and cooperative tasks (Matarić, 1994) with multiple robots, grasping with restricted search spaces (Platt et al., 2006), and mobile robot navigation (Dorigo and Colombetti, 1993).”
Growing Action Spaces (2019, background): “Such reward shaping allows learning policies in an easier MDP, which can then be transferred to the more difficult sparse-reward task (Colombetti & Dorigo, 1992; Ng et al., 1999).”
Combining manual feedback with subsequent MDP reward signals for reinforcement learning (2010, result): “Shaping rewards [4, 11] is changing the output of the reward function to learn the same task, as we did in Method 1.”
Incremental Evolution of Complex General Behavior (1997, background): “Neuro-Evolution(NE), the training of Artiicial Neural Networks with Genetic Algorithms, has proven an attractive alternative to standard learning methods in reinforcement environments (Colombetti and Dorigo 1992; Noll et al. 1990; Noll and Parisi 1995; Yamauchi and Beer 1994).”
Transfer Learning for Reinforcement Learning Domains: A Survey (2009, background): “Reward shaping(Colombetti and Dorigo, 1993; Mataric, 1994) in an RL context typically refers to allowing agent to train on an artificial reward signal rather thanR.”
Transfer Learning via Inter-Task Mappings for Temporal Difference Learning (2007, background): “Many techniques exist (Selfridge et al., 1985; Colombetti and Dorigo, 1993; Asada et al., 1994) which attempt, with more or less success, to speed up the learning process.”
Solving Non-Markovian Control Tasks with Neuro-Evolution (1999, unknown): “A number of researchers have applied task decomposition, or shaping, to make learning complex tasks tractable [Colombetti and Dorigo, 1992; Perkins and Hayes, 1996; Singh, 1992].”
Reinforcement learning from simultaneous human and MDP reward (2012, background): “In the RL literature, reward shaping adds the output of a shaping function to the original MDP reward, creating a new reward to learn from instead [3] [10].”
Morphological change in machines accelerates the evolution of robust behavior (2011, background): “It is known that environmental and social scaffolding accelerate the rate of learning in humans (21) and robots (22).”
Behavior transfer for value-function-based reinforcement learning (2005, methodology): “These benchmark players are built on the UvA Trilearn team [5] and the CMUnited-99 team [20], whereas previous publications [21] use players built on the CMUnited-99 players.”
Robust non-linear control through neuroevolution (2003, background): “A number of researchers have used shaping to make learning complex tasks tractable (Colombetti and Dorigo 1992; Jacobs et al. 1991; Perkins and Hayes 1996; Singh 1992).”
Value Functions for RL-Based Behavior Transfer: A Comparative Study (2005, background): “…manipulating the transition function, the agent's initial state, and/or the reward function , such as directed training (Selfridge, Sutton, & Barto 1985), learning from easy missions (Asada et al. 1994), and reward shaping (Colombetti & Dorigo 1993; Mataric 1994), as discussed in the Introduction.”
Transfer via inter-task mappings in policy search reinforcement learning (2007, background): “Reward shaping [2, 5] implicitly guides agents by modifying the reward function.”