Dataset Information

Learning from animals: How to Navigate Complex Terrains.

ABSTRACT: We develop a method to learn a bio-inspired motion control policy using data collected from hawkmoths navigating in a virtual forest. A Markov Decision Process (MDP) framework is introduced to model the dynamics of moths and sparse logistic regression is used to learn control policy parameters from the data. The results show that moths do not favor detailed obstacle location information in navigation, but rely heavily on optical flow. Using the policy learned from the moth data as a starting point, we propose an actor-critic learning algorithm to refine policy parameters and obtain a policy that can be used by an autonomous aerial vehicle operating in a cluttered environment. Compared with the moths' policy, the policy we obtain integrates both obstacle location and optical flow. We compare the performance of these two policies in terms of their ability to navigate in artificial forest areas. While the optimized policy can adjust its parameters to outperform the moth's policy in each different terrain, the moth's policy exhibits a high level of robustness across terrains.

SUBMITTER: Zhu H

PROVIDER: S-EPMC6952082 | biostudies-literature | 2020 Jan

REPOSITORIES: biostudies-literature

ACCESS DATA

Publications

Learning from animals: How to Navigate Complex Terrains.

Zhu Henghui H Liu Hao H Ataei Armin A Munk Yonatan Y Daniel Thomas T Paschalidis Ioannis Ch IC

PLoS computational biology 20200109 1

We develop a method to learn a bio-inspired motion control policy using data collected from hawkmoths navigating in a virtual forest. A Markov Decision Process (MDP) framework is introduced to model the dynamics of moths and sparse logistic regression is used to learn control policy parameters from the data. The results show that moths do not favor detailed obstacle location information in navigation, but rely heavily on optical flow. Using the policy learned from the moth data as a starting poi ...[more]

PMID: 31917816

Similar Datasets

Project description:This paper presents several variations of a microscale magnetic tumbling ( μ TUM) robot capable of traversing complex terrains in dry and wet environments. The robot is fabricated by photolithography techniques and consists of a polymeric body with two sections with embedded magnetic particles aligned at the ends and a middle nonmagnetic bridge section. The robot's footprint dimensions are 400 μ m × 800 μ m. Different end geometries are used to test the optimal conditions for low adhesion and increased dynamic response to an actuating external rotating magnetic field. When subjected to a magnetic field as low as 7 mT in dry conditions, this magnetic microrobot is able to operate with a tumbling locomotion mode and translate with speeds of over 60 body lengths/s (48 mm/s) in dry environments and up to 17 body lengths/s (13.6 mm/s) in wet environments. Two different tumbling modes were observed and depend on the alignment of the magnetic particles. A technique was devised to measure the magnetic particle alignment angle relative to the robot's geometry. Rotational frequency limits were observed experimentally, becoming more prohibitive as environment viscosity increases. The μ TUM's performance was studied when traversing inclined planes (up to 60°), showing promising climbing capabilities in both dry and wet conditions. Maximum open loop straight-line trajectory errors of less than 4% and 2% of the traversal distance in the vertical and horizontal directions, respectively, for the μ TUM were observed. Full directional control of μ TUM was demonstrated through the traversal of a P-shaped trajectory. Additionally, successful locomotion of the optimized μ TUM design over complex terrains was also achieved. By implementing machine vision control and/or embedding of payloads in the middle section of the robot, it is possible in the future to upgrade the current design with computer-optimized mobility through multiple environments and the ability to perform drug delivery tasks for biomedical applications.

Dataset Information

Learning from animals: How to Navigate Complex Terrains.

Publications

Learning from animals: How to Navigate Complex Terrains.

Similar Datasets

OmicsDI is part of the ELIXIR infrastructure

Tweets