Francesco Biral

Via Sommarive, 9 - 38123 Povo
tel. 0461 282513
francesco.biral[at]unitn [dot] it
SVILUPPO DI CONTROLLORI DI BASSO LIVELLO PER VEICOLI A GUIDA AUTONOMA DA COMPETIZIONE
MSc students (2020)
venerdì 07 febbraio 2020

Sviluppo di metodi innovativi di controllo della dinamica longitudinale per veicoli racing a guida autonoma.

PiĆ¹ dettagli nel documento allegato.

Full vehicle model validation for the analysis of subjective handling performances (at Siemens)
venerdì 07 febbraio 2020

MSc Thesis at Siemens Industry Software NV Digital Factory Division PLM Interleuvenlaan 68 Researchpark Haasrode Zone 1 B-3001 Leuven (Belgium). 

Details of the thesis are explained in the document attached.

Investigation of pre-crash simulations of automated vehicles (@Siemens)
venerdì 07 febbraio 2020

Msc Thesis to compare vehicle models with different levels of fidelity for pre-crash simulation in a highway scenario and in an intersection crossing scenario.

It has to be done at Siemens Industry Software NV Digital Factory Division PLM Interleuvenlaan 68 Researchpark Haasrode Zone 1 B-3001 Leuven (Belgium).

Details in the attached document

Thesis proposals related to research projects in Autonomous vehicle
venerdì 11 ottobre 2019

Vehicle dynamics modelling and indentifications

  1. Algorithms for learning and adapting vehicle dynamics

    Develop algorithms for learning longitudinal/lateral dynamics (or both) and for on-line adaptation using deterministic model and neural networks (hybrid approach) using both experimental data and simulation results.
  2. Identification of tyre characteristics from on-board experimental data

    Development and validation of algorithms for identification of tyre characteristics using simulated and experimental data to be used both off-line and on-line.
  3. Vehicle state estimation and localisation using cameras and lidars

    Development and validation of algorithms for on-line vehicle self state estimation and localisation using on-board sensors (i.e.e IMU, cameras and lidars, etc). Activity to be done using DeepLearning techniques and deployed on Nvidia Jetson TX2 on an autonomous vehicle platform.

Vehicle dynamics control

  1. Development of control strategies based on MPC and NL-MPC

    Development and validation of vehicle longitudinal and lateral control based on MPC and NL-MPC techniques to optimally track high level defined trajectories. To be tested both using simulation of numerical vehicle models and an autonomous vehicle platform.
  2. Comparison of different lateral control strategies

    Compare different lateral control strategies (from simple kinematic based up to MPC based) both using simulation of numerical vehicle models and an autonomous vehicle platform.

Motion planning

  1. Algorithm for planning overtaking with many competing vehicles

    Development and validation of algorithms for planning overtaking manoeuvres in racing environment. Different techniques will be explored and validated both using simulation of numerical vehicle models and an autonomous vehicle platform .
Physical Modeling and Implementation of a Virtual Radar Sensor and a Virtual Sonar Sensor in Midgard Engine (UE4) (@Antemotion)
MSc Mechatronics Engineering, MSc Computer science, MSc Information and telecommunication Enginnering 
venerdì 11 ottobre 2019

Autonomous Driving is an hot topic and cannot be developed only with on road experimental driving. High fidelity simulations are mandatory to test unxpected and diffcult driving situations.

To this end the simulation environement has to accurately model sensors integrated in the real vehicle. This tehsis work inc ollaboration with antemotion http://antemotion.com aim at developing a physical model of  a Radar Sensor and aVirtual Sonar in Midgard Engine (UE4) that is the basis of their driving simulator.

The two sensors in question are typically used in the field of autonomous driving, and given their particular dependence on the geometry of the surrounding environment, it is quite natural to exploit the potential of Midgard UE4 to model these sensors. Analysis of the state of the art in the field of mathematics of lenses and the reflection lighting model of Radar. Definition of the specifications for an Radar model  based on the physics of materials, exploiting the potential of Midgard UE4 (PBR, Raytracing etc.) Implementation of sensors.

(optional) Validation of the Radar / SOnar  model - this activity depends on the availability of some of our partners.

Physical Modeling and Implementation of a Virtual Lidar Sensor and a Virtual RGB Camera in Midgard Engine (UE4) (@antemotion)
MSc Mechatronics Enginneing, MSc Computer science, MSc Information and telecommunication Enginnering 
venerdì 11 ottobre 2019

Autonomous Driving is an hot topic and cannot be developed only with on road experimental driving. High fidelity simulations are mandatory to test unxpected and diffcult driving situations.

To this end the simulation environement has to accurately model sensors integrated in the real vehicle. This tehsis work inc ollaboration with antemotion http://antemotion.com aim at developing a physical model of  a Lidar Sensor and a RGB Camera in Midgard Engine (UE4) that is the basis of their driving simulator.

The two sensors in question are typically used in the field of autonomous driving, and given their particular dependence on the geometry of the surrounding environment, it is quite natural to exploit the potential of Midgard UE4 to model these sensors. Analysis of the state of the art in the field of mathematics of lenses and the reflection lighting model of LIDAR. Definition of the specifications for an RGB / Monochrome camera model with distortions introduced by the lens to be introduced in Midgard UE4Definition of the specifications for a LIDAR model with lighting and reflection model based on the physics of materials, exploiting the potential of Midgard UE4 (PBR, Raytracing etc.) Implementation of sensors

(optional) Validation of the LIDAR / CAMERA model - this activity depends on the availability of some of our partners.

Thesis proposals in Biomechanics and human motion analysis
venerdì 11 ottobre 2019

Human physiological models

  1. Physiological model to quantify oxygen consumption and lactate production

    Develop hybrid (model based plus neural networks)models oxygen consumption and lactate production to accuratley quantify energy expediture during physical exercise. Model will be validated on in-door and outdoor experimental data of non-professional and professional cyclists.

Biomechanical models

  1. Biomechanical model of walking person

    Develop symbolic models of a walking person for fast and efficient numerical simulation. Models should be compare with openSim software results and experimental data from motion capture system and other physiological data.
  2. Biomechanical model of cyclist

    Develop symbolic models of a cyclist with main muscles for fast and efficient numerical simulation. Models should be compare with openSim software results and experimental data from motion capture system and other physiological data.

Human motion and performance analysis

  1. Analysis of cycling performance using Optimal Control Approach

    Use optimal control method to study the performance humans for cycling task. Results should be compared with experimental data from indoor and outdoor tests.
  2. Analysis of walking performance using Optimal Control Approach

    Use optimal control method to study the performance humans for walking task. Results should be compared with experimental data from indoor tests.
  3. Analysis of running performance using Optimal Control Approach

    Use optimal control method to study the performance humans for running task. Results should be compared with experimental data from indoor tests.
Neural network training using synthetic image from high resolution driving simulator (@Antemotion)
MSc Mechatronics Engineering, MSc Computer Science, MSc Information and Communication Engineering 
venerdì 11 ottobre 2019

Nvidia XAVIER integrated in a RealTime driving simulator for fast ADAS prototyping: Implementing on an FPGA the CSI2 protocol to feed a virtual camera view from the render engine directly onto the Nvidia Xavier via the standard camera connector (CSI2 protocol)

The thesis work will be done @Antemotion (https://antemotion.com) in Rovereto.