Blätter-Navigation

An­ge­bot 128 von 316 vom 13.06.2018, 11:58

logo

BASELABS GmbH

The car of the future drives auto­mat­ic­ally. BASELABS provides the soft­ware required for the devel­op­ment of autonom­ous vehicles and mod­ern driver assist­ance sys­tems.
BASELABS is grow­ing - are you grow­ing with us?

Work­ing Stu­dent (m/w) / Thesis Topic: Imple­ment­a­tion and Eval­u­ation of a Dynamic Occu­pancy Grid Algorithm using an Object-Track­ing Frame­work

Work­ing field:

BASELABS devel­ops data-fusion and object-track­ing algorithms in the auto­mot­ive domain. Cur­rently, the com­pany designs a new frame­work (BASELABS Cre­ate Embed­ded) for data-fusion sys­tems that are ready for deploy­ment in safety-crit­ical applic­a­tions. The frame­work is imple­men­ted in C and a spe­cial­ized con­fig­ur­a­tion lan­guage called Trait-C. Trait-C adds advanced fea­tures such as type-safe tem­plates and com­pile-time poly­morph­ism to C.
Grid map­pings, or grid-based estim­a­tion algorithms are well-known algorithms which have been suc­cess­fully used for dec­ades in robot­ics. One applic­a­tion of this algorithm is the estim­a­tion of an occu­pancy grid [1]. The main lim­it­a­tion of this approach is its applic­ab­il­ity for static envir­on­ments only. With a con­sid­er­able num­ber of dynamic objects in the envir­on­ment the qual­ity of the estim­ated world
model sig­ni­fic­antly deteri­or­ates. Recently, Multi-Instance Bernoulli PHD fil­ters (aka Dynamic Grids) and sim­ilar tech­niques have been pro­posed as altern­at­ive approaches for envir­on­mental model estim­a­tion suit­able for highly dynam­ical scen­arios. These meth­ods rep­res­ent the world around an object as a set of mov­ing particles, where the lat­ter are employed both for the objects’ state and dens­ity-occu­pancy estim­a­tion. The estim­a­tion res­ults for static and dynamic envir­on­ments can used as an input for higher-level func­tion­al­ity such as object clas­si­fic­a­tion using machine learn­ing algorithms or advanced driver-assist­ance sys­tems.
The major object­ive of this task is to con­duct a feas­ib­il­ity study on a pos­sible exten­sion of BASELABS Cre­ate Embed­ded for GPGPU hard­ware (e.g., by Nvidia) with a pro­to­typ­ical imple­ment­a­tion of a grid-based estim­a­tion algorithm along with its optim­iz­a­tion for exe­cu­tion on GPGPU hard­ware and/or embed­ded devices with par­al­lel cap­ab­il­it­ies. The basic algorithm which is inten­ded to be imple­men­ted
is described in pseudo-code in [2]. The study there­fore invest­ig­ates the gen­eral pos­sib­il­ity of extend­ing BASELABS Cre­ate Embed­ded and the Trait-C lan­guage to sup­port algorithms like those presen­ted in [2].

THE INDI­VIDUAL TASKS ARE AS FOL­LOWS

1. Pro­to­typ­ic­ally imple­ment the algorithm in BASELABS Cre­ate Embed­ded using C and Trait-C. This includes:

  • adapt­ing the frame­work and/or the lan­guage to effect­ively sup­port [1] if neces­sary,
  • imple­ment­ing sev­eral basic blocks from the paper [2] for par­al­lel archi­tec­tures using CUDA and OpenCL. The com­pany can provide sev­eral ref­er­ence imple­ment­a­tions for key steps which may be needed to be adap­ted or changed.

2. Bench­mark the par­al­lel imple­ment­a­tions against each other and a sequen­tial imple­ment­a­tion.

3. Invest­ig­ate pos­sible implic­a­tions for applic­a­tions that use MISRA-C con­straints as typ­ical in the auto­mot­ive domain.

[1] https://en.wikipedia.org/wiki/Occupancy_grid_mapping
[2] https://arxiv.org/abs/1605.02406

Require­ments:

  • Stu­dent in a Com­puter Sci­ence (or related) Mas­ter’s or Dip­loma pro­gram at TU Dresden
  • Good Bach­elor or “Vordip­lom” in Com­puter Sci­ence (or related sub­ject)
  • Know­ledge in C, C++ or C#
  • Desir­able is exper­i­ence with com­pilers or pro­gram­ming GPU’s

The work­ing stu­dent job requires per­sonal attend­ance in our Chem­nitz office for 2-3 days / week.
The thesis will be super­vised aca­dem­ic­ally at the chair for com­piler con­struc­tion at TU Dresden.

What we of­fer:

  • Var­ied and chal­len­ging tasks at the Chem­nitz loc­a­tion
  • Work­ing in and with a team of experts
  • Appre­ci­at­ive work envir­on­ment includ­ing free­dom for con­tri­bu­tion of own ideas and solu­tions
  • Fam­ily-friendly work­ing con­di­tions and flex­ible work­ing hours
  • Free fruit and drinks
  • Par­ti­cip­a­tion in team events

How to ap­ply:

Please apply online: http://bit.ly/BASELABS

Your con­tact is Dr. Sven Karol.