Robots that can adapt like animals

  title={Robots that can adapt like animals},
  author={Antoine Cully and Jeff Clune and Danesh Tarapore and Jean-Baptiste Mouret},
Robots have transformed many industries, most notably manufacturing, and have the power to deliver tremendous benefits to society, such as in search and rescue, disaster response, health care and transportation. They are also invaluable tools for scientific exploration in environments inaccessible to humans, from distant planets to deep oceans. A major obstacle to their widespread adoption in more complex environments outside factories is their fragility. Whereas animals can quickly adapt to… 

Creative Adaptation through Learning

This work aims to provide the algorithmic foundations that will allow physical robots to be more robust, effective and autonomous through three contributions: the behavioral repertoires, the damage recovery using these repertoires and the transfer of knowledge across tasks.

Evolving Robust Robots

Investigating the intrinsic robustness of NEAT evolved networks found that adding noise to simulation either did not affect or improved outcomes, so designing for the reality gap problem and for robustness are not at cross purposes in this domain.

Artificial intelligence: Robots with instincts

An evolutionary algorithm has been developed that allows robots to adapt to unforeseen change and enables damaged robots to quickly regain their ability to perform tasks.

A brittle star-like robot capable of immediately adapting to unexpected physical damage

It is found that physical interaction between arms plays an essential role for the resilient inter-arm coordination of brittle stars and this finding will help develop resilient robots that can work in inhospitable environments.

Damage Recovery for Simulated Modular Robots Through Joint Evolution of Morphologies and Controllers

A new approach is introduced that generates resilient artificial modular robots by evolving the robot morphology along with its controller and demonstrates that during evaluation, when robots are deliberately faced to motor failures, the evolution process can optimize and generate new morphologies for which the robot’s behavior is less affected by damage.

Continuous learning of emergent behavior in robotic matter

It is shown that the physical connection between the units is enough to achieve learning, and no additional communication or centralized information is required, and such a distributed learning approach can be easily scaled to larger assemblies.

Online Damage Recovery for Physical Robots with Hierarchical Quality-Diversity

The Hierarchical Trial and Error algorithm, which uses a hierarchical behavioural repertoire to learn diverse skills and leverages them to make the robot adapt quickly in the physical world, and shows that the hierarchical decomposition of skills enables the robot to learn more complex behaviours while keeping the learning of the repertoire tractable.

Reset-free Trial-and-Error Learning for Robot Damage Recovery

Behavioral Repertoires for Soft Tensegrity Robots

This work employs a Quality Diversity Algorithm running model-free on a physical soft tensegrity robot that autonomously generates a behavioral repertoire with no a priori knowledge of the robot’s dynamics, and minimal human intervention to provide a road map for increasing the behavioral capabilities of mobile soft robots through real-world automation.

Automating the Incremental Evolution of Controllers for Physical Robots

This article shows that it is possible, for the first time, to incrementally evolve a neural robot controller for different obstacle avoidance tasks with no human intervention and offers a high level of robustness and precision that could potentially open up the range of problems amenable to embodied evolution.



Fast damage recovery in robotics with the T-resilience algorithm

The T-resilience algorithm is introduced, a new algorithm that allows robots to quickly and autonomously discover compensatory behavior in unanticipated situations and consistently leads to substantially better results than the other approaches.

On autonomous robots

  • G. Bekey
  • Computer Science
    The Knowledge Engineering Review
  • 1998
Autonomous robots are the intelligent agents par excellence in that robots are embodied agents, situated in the real world, subject both to the joys and sorrows of the world and to its physical laws.

Innately adaptive robotics through embodied evolution

This work investigates autonomous robot adaptation, focussing on damage recovery and adaptation to unknown environments, and introduces an embodied evolutionary algorithm, shown to be able to control the motion of a robot snake effectively and inherently recovers the snake’s motion after damage.

Self-organized adaptation of a simple neural circuit enables complex robot behaviour

This paper presents a meta-analyses of the response of the immune system to solar energy perturbation using a probabilistic approach called “computational neuroscience”.

Evolving coordinated quadruped gaits with the HyperNEAT generative encoding

It is demonstrated that HyperNEAT, a new and promising generative encoding for evolving neural networks, can evolve quadruped gaits without an engineer manually decomposing the problem.

Evolving Gaits for Physical Robots with the HyperNEAT Generative Encoding: The Benefits of Simulation

This paper tested the hypothesis that the beneficial properties of Hyper NEAT would outperform the simpler encoding if HyperNEAT gaits are first evolved in simulation before being transferred to reality, and it was confirmed, resulting in the fastest gaits yet observed for this robot.

Trial by fire [rescue robots]

  • R. Murphy
  • Materials Science
    IEEE Robotics & Automation Magazine
  • 2004
An overview of the use of robots for USAR is provided, concentrating on what robots were actually used and why, and the roles that the robots played in the response and the impact of the physical environment on the platforms.

Reinforcement learning in robotics: A survey

This article attempts to strengthen the links between the two research communities by providing a survey of work in reinforcement learning for behavior generation in robots by highlighting both key challenges in robot reinforcement learning as well as notable successes.

A Biologically Inspired Controller for Hexapod Walking: Simple Solutions by Exploiting Physical Properties

  • J. SchmitzJ. DeanThomas KindermannMichael SchummHolk Cruse
  • Biology
    The Biological Bulletin
  • 2001
Investigations of the stick insect Carausius morosus indicate that these animals gain their adaptivity and flexibility mainly from the extremely decentralized organization of the control system that generates the leg movements.

Locomotion Analysis of Hexapod Robot

The most studied problem for multi-legged robots concerns how to determine the best sequence for lifting off and placing the feet (gait/locomotion planning).