caor@mines-paristech.fr
26
Jul

Sensor fusion for unsupervised learning of rare cases in autonomous vehicles

Thesis Title:

Sensor fusion for unsupervised learning of rare cases autonomous vehicles

Background:

Huawei is working on key components of L2-L3 autonomous driving platform and progressively shifting focus to development of breakthrough technologies required for L4-L5 autonomy. Tomorrow self-driving cars powered by AI will combine edge and cloud compute with vast number of sensors to safely and autonomously drive customers and deliver merchandise. At Huawei we develop a full-stack of technologies to enable this dream, including compute units, sensors, communication and cloud. We are seeking the best candidates for PhD CIFRE with a background in computer vision, deep learning, reinforcement learning, mapping, perception, sensor fusion, cognition and other related areas, to work as a part of IoV team in Paris Research Center (PRC). As a member IoV PRC you will closely work with multiple teams worldwide to grow your expertise and successfully transfer your research results into real products.

Research topic:

The core subject of this PhD should focus on solving the problem of sensor fusion for unsupervised learning of rare cases by leveraging vast amount of data available using multiple sensors and vehicles. By combining results of processing of the same physical entity by multiple redundant sensors and map data one could detect failures and root cause of it by comparing results among multiple sources. Unsupervised learning should further improve accuracy of finding errors especially in case of rare events, where training data for conventional sensors is absent. This data doesn’t have manual annotations, however automated SOTA methods could be used to extract valuable information about perception, localization, mapping to turn the problem into tractable. Unsupervised learning is considered to be among the most challenging problems as of today, therefore we expect to see tremendous progress in this area that could enable highly safe automated vehicles to become a part of our daily life.

Description of research activities:

  • Study state of the art on perception, sensor fusion and cognition
  • Study state of the art of supervised/unsupervised learning
  • Identify major bottlenecks in SOTA of (1 and 2) with application to self-driving car problems
  • Propose a new solution to one of the main bottlenecks (3) with focus on large scale practical application for self-driving cars
  • Research and develop algorithm based on the proposed solution (4) that could identify rare case failures of various sensors in unsupervised manner within focus of improving accuracy of sensor data for self-driving cars
  • Apply proposed algorithm (5) to the domain of self-driving cars using existing or specifically collected datasets
  • Publish research results in top conferences and participate to scientific seminars

Supervision:

This PhD will be supervised jointly between Huawei Technologies France and the Robotics Centre of Mines ParisTech and Armines

Prerequisites:

The candidate should be motivated to carry out world class research and should have a Master in Computer Vision and/or Robotics. He/She should have solid skills in the following domains:

  • Implement Code in Python & C++
  • Apply or use existing libraries for deep learning in project related tasks
  • Good knowledge in Git, ROS, OpenCV, Boost, multi-threading, CMake, Make and Linux systems
  • Code and algorithm documentation
  • Project reporting and planning
  • Writing of scientific publications and participation in conferences
  • Fluency in spoken and written English; French and/or Chinese is a plus
  • Intercultural and coordination skills, hands-on and can-do attitude
  • Interpersonal skills, team spirit and independent working style

Contact:

Interested candidates must send a detailed CV including their Master’s scores to dzmitry.tsishkou@huawei.com and bogdan.stanciulescu@mines-paristech.fr

Candidates must hold a working visa in France, or be nationals of one of the states of the European Union or the European Economic Area.

Deadline:
We aim to fill this position as soon as possible with the aim to commence in 2nd half 2019. Applications will be considered until a suitable candidate has been found.

Funding and location:
The PhD will be funded by a CIFRE contract and hosted in Paris area (Boulogne Billancourt and Paris), France.

26
Jul

Realistic modelling of driving scenarios based on sensor fusion for autonomous cars

Thesis Title:

Realistic modelling of driving scenarios based on sensor fusion for autonomous cars

Background:

Huawei is working on key components of L2-L3 autonomous driving platform and progressively shifting focus to development of breakthrough technologies required for L4-L5 autonomy. Tomorrow self-driving cars powered by AI will combine edge and cloud compute with vast number of sensors to safely and autonomously drive customers and deliver merchandise. At Huawei we develop a full-stack of technologies to enable this dream, including compute units, sensors, communication and cloud. We are seeking the best candidates for PhD CIFRE with a background in computer vision, deep learning, reinforcement learning, mapping, perception, sensor fusion, cognition and other related areas, to work as a part of IoV team in Paris Research Center (PRC). As a member IoV PRC you will closely work with multiple teams worldwide to grow your expertise and successfully transfer your research results into real products.

Research topic:

The core subject of this PhD should focus on solving the problem of realistic modelling of driving scenarios based on sensor fusion by leveraging vast amount of data available using multiple sensors and vehicles. This data doesn’t have manual annotations, however automated SOTA methods could be used to extract valuable information about perception, localization, mapping to turn the problem into tractable. Simulation is playing major role in increased accuracy of perception, sensor fusion, mapping and planning and control. However, at present there is no realistic model of driving scenarios that could adequately imitate rational behavior of movable objects. That restricts the scope of usage of simulation, so at present we are still required to collect large amount of real world data. By introducing realistic driving models one could close the gap and accelerate development of self-driving cars by increasing value of simulation. Realistic modelling of driving scenarios is considered to be among the most challenging problems as of today, therefore we expect to see tremendous progress in this area that could enable highly safe automated vehicles to become a part of our daily life.

Description of research activities:

  • Study state of the art on perception, sensor fusion and cognition
  • Study state of the art of modeling of driving scenarios and planning/control
  • Identify major bottlenecks in SOTA of (1 and 2) with application to self-driving car problems
  • Propose a new solution to one of the main bottlenecks (3) with focus on large scale practical application for self-driving cars
  • Research and develop algorithm based on the proposed solution (4) that could model realistic driving scenarios with focus on improving self-driving planning/control
  • Apply proposed algorithm (5) to the domain of self-driving cars using existing or specifically collected datasets
  • Publish research results in top conferences and participate to scientific seminars

Supervision:

This PhD will be supervised jointly between Huawei Technologies France and the Robotics Centre of Mines ParisTech and Armines

Prerequisites:

The candidate should be motivated to carry out world class research and should have a Master in Computer Vision and/or Robotics. He/She should have solid skills in the following domains:

  • Implement Code in Python & C++
  • Apply or use existing libraries for deep learning in project related tasks
  • Good knowledge in Git, ROS, OpenCV, Boost, multi-threading, CMake, Make and Linux systems
  • Code and algorithm documentation
  • Project reporting and planning
  • Writing of scientific publications and participation in conferences
  • Fluency in spoken and written English; French and/or Chinese is a plus
  • Intercultural and coordination skills, hands-on and can-do attitude
  • Interpersonal skills, team spirit and independent working style

Contact:

Interested candidates must send a detailed CV including their Master’s scores to dzmitry.tsishkou@huawei.com and bogdan.stanciulescu@mines-paristech.fr

Candidates must hold a working visa in France, or be nationals of one of the states of the European Union or the European Economic Area.

Deadline:

We aim to fill this position as soon as possible with the aim to commence in 2nd half 2019. Applications will be considered until a suitable candidate has been found.

Funding and location:

The PhD will be funded by a CIFRE contract and hosted in Paris area (Boulogne Billancourt and Paris), France.

13
Jun

Car Onboard Visualization of Egocentric Real-Time 3D Reconstruction of Road Scene

PhD position offering

Position and duration: PhD Thesis – 3 Years full time contract

Starting date: 1st of October 2016

Context :

This PhD thesis Subject is a part of an ongoing work between Mines ParisTech and PSA Peugeot Citroën Group, focused on car security. The context is the use of onboard visual digital information to inform the driver and increase his security. The digital information is acquired outside of the vehicle through the use of sensors. This is a confidential work, potentially leading to a patent, because of this, the scenario of use cannot be precisely described in this subject.

Objectives

There are two objectives in the thesis : 1) Reconstruct a real-time 3D view of the car environment, acquired from a fixed point of view on a moving car, and 2) Master the visual perception of that 3D data by a human, positioned within the car, and ensure the pertinence of its use for a given set of scenarios.

The point of view for the 3D reconstruction does not have to be a very wide angle view, but the visual refreshment of the 3D data has to deal with the specific constraints of car driving. Indeed, for high speeds, the processing time and delays can have a negative effect on the visual data presented to the driver and the algorithms for 3D reconstruction should account for spatial shifting effects.

One of the main purposes of this thesis is to perform a realistic but partial 3D environment reconstruction of the road environment, given the inherent constraints of the automotive field: speed of the vehicle, road visibility, etc. The aim is not in performing a complete 3D scene reconstruction, from a single viewpoint, but a real-time partial reconstruction with respect to the speed of the vehicle and the inherent perturbation phenomena: change in vehicle’s direction and other secondary motions. If a certain speed threshold is overpassed, unwanted effects can occur such as the processing time and the information rendering, therefore some strategies of image acquisition should be studied, in order to compensate these inconvenient situations. In a second phase, the emerging algorithms and technology should be validated from the perceptive point of view, by user experience tests. Use cases and tests will be defined for the system’s evaluation.

Description of work:

After a preliminary phase of the field’ state of the art, the PhD candidate should propose pertinent technical solutions of innovations over the existing technologies, able to respond to the challenges risen by the state of art. The second phase will be the algorithms implementation phase on the vehicle platform.

At the end, a definition phase of the realistic use case from the safety and security point of view will be analyzed, for the system assessment.

Finally, realistic use cases will be set up and evaluated. One central question as for the perceptive feasibility of the concept, is the question of visual behavior of the user. Indeed, it is unknown how the users will deal with the presence of visual information within the car, that is relative to events and objects that are outside of the car. Especially we will focus on accommodative behaviors, indeed, users may have two accommodative strategies: 1) switch between interior/exterior accommodation points or 2) keep an -outside of the car- accommodation strategy and use the digital visual information available within the car in their peripheral visual field. The open question being which, if any, one of these strategies is actually viable in the proposed interface.

Candidate profile

MSc Degree in Electrical Engineering, Computer Science or Physics.

Requirements:

Scientific knowledge in Virtual or Augmented Reality, Computer Vision and real-time computer programming.

  • Job information:

The successful candidate will be enrolled in the Mines-ParisTech doctoral program and employee of the PSA Peugeot-Citroën Company.

  • Working office/Laboratory :

Centre de Robotique / Mines ParisTech

60 boulevard Saint Michel

75272 Paris Cedex 06

http://caor-mines-paristech.fr/en/home/

  • PhD Supervisors:

Director: Philippe Fuchs: philippe.fuchs@mines-paristech.fr

Scientific supervisor: Bogdan Stanciulescu: bogdan.stanciulescu@mines-paristech.fr

Scientific supervisor: Alexis Paljic: alexis.paljic@mines-paristech.fr

Administrative referent:

Mme Christine Vignaud

Phone: +33-(0)1 40 51 92 55

Email: christine.vignaud@mines-paristech.fr