Teaching robots to avoid moving obstacles

With self-driving cars being tested on the streets, automated machines in factories, indoor robots in offices, and autonomous systems to help persons with disability — we’re heading for a world where machine is replacing man in more ways than we can imagine.

How to operate robots reliably in a dynamic environment has, understandably, been a subject of much research over the last decade. One such researcher is Anindya Harchowdhury with the IITB-Monash Research Academy, who is working on a project titled, ‘Autonomous mapping in the presence of dynamic obstacles using multiple sensors’.

Fig. 1: Self-driving car conceptual model, (wiki image courtesy)

“Probabilistic analysis of sensory perception and mapping has brought in far more accuracy in mapping and the ability to control a robot in an unknown environment,” explains Anindya. “However, classification of obstacles based on their dynamic behaviour has not been studied much. Especially, robots operating in a low dynamic environment, which encounter significant challenges in obstacle identification due to infrequent mobility of the obstacles.”

Localisation and mapping in a dynamic environment is a key problem in robotics. The characterisation of a robot’s ability to detect obstacles with different dynamic properties is essential to understand the optimal path planning scheme for the robot.

Anindya and his colleagues are trying to develop a novel technique to help detect dynamic obstacles, and hope to build a library of common obstacles that would help a robot to take decisions faster. While 3D perception brings in much better visualisation of an object compared to 2D, it also enhances the cost of the robot significantly. Anindya is therefore attempting to design a 3D range sensor using a 2D one, which would reduce the cost of 3D perception without compromising on quality. “The 3D range sensor that we are designing will provide a user much more flexibility to choose various parameters to perceive the objects with higher resolution at different distances. It will also offer optimality in scanning both in space and time,” he says.

Anindya envisages that his research may be applied to self-driving cars in the future. “Autonomous cars will not only bring in a revolution in the automobile sector, but also play a big role in allowing physically challenged people to move more freely. Our work on the learning mechanism of dynamic obstacles will help build self-driving vehicles operable in environments equipped with objects of varieties of dynamic behaviour.”

The IITB-Monash Research Academy is a collaboration between India and Australia that endeavours to strengthen scientific relationships between the two countries. Graduate research scholars like Anindya study for a dually-badged PhD from both IIT Bombay and Monash University, spending time at both institutions to enrich their research experience.

Says Prof Murali Sastry, CEO of the Academy, “The map is a representation of an environment. Building a map using an autonomous mobile robot involves challenges, particularly when the information about the environment is unknown. The map must be consistent and complete. Anindya is confident that his prototype can be used both indoors and outdoors. This project can go a long way in ensuring that we will see self-driving cars on our roads in the foreseeable future.”

Research scholar: Anindya Harchowdhury, IITB-Monash Research Academy

Project title: Autonomous mapping in the presence of dynamic obstacles using multiple sensors

Supervisors: Prof Leena Vachhani and Prof Lindsay Kleeman

Contact details: anindya.harchowdhury@monash.edu

This story was written by Mr Krishna Warrier based on inputs from the research student, his supervisors and IITB-Monash Research Academy. Copyright IITB-Monash Research Academy.