Rodrigues, Maradona (2018) Human-like motion planning for autonomous ground vehicle applications. EngD thesis, University of Warwick.

PDF WRAP_Theses_Rodrigues_2018.pdf - Submitted Version Embargoed item. Restricted access to Repository staff only until 10 June 2022. Contact author directly, specifying your specific needs. - Requires a PDF viewer. Download (12Mb)

Request Changes to record.

Abstract

To overcome the existing challenges within the passenger car sector such as improving motion safety, reducing traffic congestion and meeting the increasing expectation of drive-comfort, the Intelligent Transportation System community have long envisioned developing autonomous vehicles capable of driving themselves without the need for human intervention. Although the technological progress and drive from the ITS community (academia, OEMs, Govt. etc) has manifested in autonomous vehicle prototypes accumulating millions of autonomously driven miles, critical technological limitations still exist preventing the mass application of the technology on public roads.

One important technological limitation of the existing autonomous systems is the inability to successfully negotiate interaction-dependent urban driving scenarios, highlighted by the number of reported collisions/ near misses and autonomous vehicle disengagements during testing. Therefore, most autonomous vehicle testing has been generally restricted to simple road geometries and less dynamic or controlled driving scenarios. There is lack of significant evidence of attempts made to demonstrate autonomous navigation of complex, ambiguous and interaction-dependent scenarios within urban environments, such as non-signalised junctions, shared crossing zones etc. While some human drivers do demonstrate the expert ability of negotiating such scenarios every day, there is a great degree of inconsistency among the human driving populace. This inconsistency with motion behaviour adaptation and decision-making in interaction-dependent scenarios, leads to unsafe, inefficient and uncomfortable driving experience. Successful autonomous vehicle motion planning in such scenarios therefore necessitates inheriting the adaptive behaviour planning with naturalistic manoeuvres and tactical decision-making abilities analogous to “expert” human drivers.

This research proposed a novel “human-like” motion planning approach with the characteristics of adaptive motion planning through “naturalistic” trajectory generation and tactical decision-making, The two foremost contribution of this research are the motion planning system framework (HAPS), that enables hybrid forms of decision-making in autonomous vehicle and an integrated local motion planning system (ATBP), which combines the behaviour and trajectory planning system to achieve the desired characteristics of expert human driving. With the proposed approach, the autonomous vehicle was shown to be superior at negotiating interaction-dependent scenarios by outperforming human drivers on the objectives of motion safety (avoid collision and near misses), motion efficiency (reduce navigation time) and motion comfort (maintain accelerations within acceptable limits) in two simulator studies. Furthermore, the application of innovation was demonstrated through the successful implementation and testing of the motion planning system on a real vehicle platform. The autonomous vehicle demonstrated the expert human-like ability to adapt its motion behaviours to firstly negotiate a selected list of highly dynamic driving scenarios in controlled environment and then, drove autonomously in un-controlled free flowing traffic in the first of its kind autonomous demonstrations in two cities in the UK.

Item Type:	Thesis or Dissertation (EngD)
Subjects:	T Technology > TE Highway engineering. Roads and pavements T Technology > TL Motor vehicles. Aeronautics. Astronautics
Library of Congress Subject Headings (LCSH):	Intelligent transportation systems, Intelligent transportation systems -- Data processing, Automated vehicles, Automated vehicles -- Computer programs, Autonomous vehicles -- Collision avoidance, Automobile driving -- Human factors
Official Date:	December 2018
Dates:	Date Event December 2018 UNSPECIFIED
Institution:	University of Warwick
Theses Department:	Warwick Manufacturing Group
Thesis Type:	EngD
Publication Status:	Unpublished
Supervisor(s)/Advisor:	McGordon, Andrew ; Marco, James
Sponsors:	Tata Motors ; Engineering and Physical Sciences Research Council
Format of File:	pdf
Extent:	xv, 215 leaves : illustrations
Language:	eng

Request changes or add full text files to a record

Repository staff actions (login required)

View Item