Intelligent Transport Lab Center for Cognitive Modeling MIPT
Internship
About Lab
The Intelligent Transport Laboratory was created in 2019 at the Moscow Institute of Physics and Technology (MIPT) on the initiative of the Scientific and Design Bureau of Computer Systems (NKB VS). Together with the Laboratory of Cognitive Dynamic Systems, it is part of the Center for Cognitive Modeling MIPT.
The laboratory specializes in the design and creation of embedded integrated electronic control systems for various types of transport (including unmanned vehicles) and intelligent image processing and synthesis systems using deep machine learning methods, Big Data technologies, virtual and augmented reality, autonomous navigation and mapping (SLAM) methods. ) and etc.
Dmitry Yudin
Head of the Intelligent Transport Lab MIPT - NKB VS
candidate of technical sciences, Associate Professor, Department of System Research, MIPT, Senior Researcher, Laboratory of Cognitive Dynamical Systems
Research interests: Technical vision Artificial Intelligence Deep learning Robotics
Main research topics associated with the development of mathematical and software for intelligent transport systems:
Onboard multispectral vision systems;
Navigation systems;
Autonomous traffic control systems;
Geographic Information Systems;
Human-machine control interfaces;
Virtual computer simulation
External environment for practicing, training and
Verification of control algorithms;
Integrated on-board information and control systems.
Applied tools Software development is carried out using Python and C++ programming languages, NVidia Cuda and TensorRT libraries, ROS robots operating system, Docker. Many projects are related to integration into embedded systems, in particular with the use of NVidia Jetson Xavier, TX2 and Nano. Tensorflow 2.0 and PyTorch are used as basic software platforms for researching deep learning algorithms.
Research and development directions
Road scene recognition
Detection, segmentation and classification of road users (cars / trucks, buses / trams, motorcyclists / cyclists, pedestrians) and road infrastructure objects (traffic lights, road signs, roadways, road markings)
Localization and mapping
Simultaneous localization and environment mapping (SLAM) based on the integration of data from vision sensors and navigation systems, including the use of deep learning
Self-driving car software
Software for a prototype of an unmanned vehicle created by the AUTO-RTK consortium
Mobile Robot Software
Software for a prototype mobile robot capable of performing courier delivery functions
Re-identification of objects, tracking and trajectory prediction
Identification, tracking and prediction of the trajectory of motion of dynamic objects based on the integration of data received from various on-board vision systems of a mobile robot
Place recognition
Solving Place recognition problems (including global localization) using neural networks based both on onboard camera images and orthomosaics. This additionally uses a three-dimensional depth map and/or semantic image segmentation.
Planning the route and behavior of the mobile robot
Solving the problems of automatic planning of the route and trajectory of the mobile robot and the implementation of its behavioral logic for various scenarios