Skip to Main content Skip to Navigation
Theses

Automatic design of behavioural models for bio-hybrid systems of animals and robots

Abstract : Robots are increasingly used in behavioural research as tools that can induce multimodal stimuli to interact with animals. However, the robots used in most studies are simple entities without the ability to make their own decisions and can only be used in simple short-lived interactions with the animals. This approach can be extended further by using autonomous and biomimetic robots able to socially integrate groups of animals with closed-loop of interactions: the robots influence the animals, which, in turn, influence the behaviour of the robots. However, the design of these robots is challenging, as they have to be accepted by the animals as social companions and respond accordingly to social and environmental stimuli. This thesis addresses the problem of automatically designing animal behavioural models, that are implemented as controllers to drive biomimetic robots into socially integrating animal groups, and forming mixed-groups of animals and robots. In particular, we investigate the integration of robots driven by biomimetic behavioural models into groups of zebrafish Danio rerio) and into groups of cockroaches (P. americana and B. germanica). In both cases, we show that these models can be automatically calibrated to drive the robots into integrating groups of animals, mimicking their behaviour and exhibiting similar collective dynamics compared to animal-only groups. The main contributions of this thesis are: 1) a methodology to automatically design microscopic models of agents behaviour to correspond to the dynamics exhibited by a macroscopic model of behaviour. This allows the automatic transition from a macroscopic description to a microscopic one. This methodology was then used to design the controllers of robots driven to socially integrate groups of cockroaches. 2) a methodology to use automatically designed behavioural models to drive socially integrated robots into modulating the collective behaviour of mixed-groups of cockroaches and robots. 3) a behavioural model that captures zebrafish behaviour individual and collective behaviour in a fragmented environment (i.e. environment with topological discontinuities). 4) a methodology to calibrate this model to exhibit the same individual and collective behaviour as the animals. We describe a set of metrics used to assess the biomimetism of this calibrated model. 5) a methodology to create mixed-groups of fish and robot. We used this calibrated model to drive robots to socially integrate groups of zebrafish.
Complete list of metadatas

Cited literature [224 references]  Display  Hide  Download

https://tel.archives-ouvertes.fr/tel-02290801
Contributor : Leo Cazenille <>
Submitted on : Wednesday, September 18, 2019 - 4:28:51 AM
Last modification on : Monday, May 4, 2020 - 11:39:32 AM
Long-term archiving on: : Saturday, February 8, 2020 - 9:07:11 PM

File

manuscript.pdf
Files produced by the author(s)

Identifiers

  • HAL Id : tel-02290801, version 1

Citation

Leo Cazenille. Automatic design of behavioural models for bio-hybrid systems of animals and robots. Artificial Intelligence [cs.AI]. Université Paris Diderot, 2018. English. ⟨tel-02290801⟩

Share

Metrics

Record views

109

Files downloads

42