notesum.ai

Published at December 9

Parameter Adjustments in POMDP-Based Trajectory Planning for Unsignalized Intersections

cs.RO

Released Date: December 9, 2024

Authors: Adam Kollarčík adn Zdeněk Hanzálek

Arxiv: http://arxiv.org/pdf/2412.06405v1

Refer to caption

	Description	Value
$A$	action set	$\left\{-2,-1,0,1\right\}$ $\mathrm{m}\text{\,}{\mathrm{s}}^{-2}$
$c$	UCB parameter	20000
$N$	opt. horizon	5
$n_{\mathrm{ep}}$	num. of episodes	3000
$n_{\mathrm{par}}$	num. of particles	300
$\gamma$	discount fact.	1
$\bm{Q}$	dyn. covariance	$\bm{0}$
$\bm{R}$	obs. covariance	$\mathrm{diag}(10^{-2},10^{-2},0)$
$\sigma_{\omega_{1}}$	IDM st. dev.	1 $\mathrm{m}\text{\,}{\mathrm{s}}^{-2}$
$\tau$	time headway	2 $\mathrm{s}$
$\delta$	acc. exponent	4
$d_{\mathrm{min}}$	IDM min. dist.	1 $\mathrm{m}$
$R_{\mathrm{acc}}$	acc. reward coef.	1
$R_{\mathrm{vel}}$	vel. reward coef.	100
$R_{\mathrm{crash}}$	crash reward	10000
$a_{\textrm{lat},\textrm{max}}$	max. lat. acc.	0.5 $\mathrm{m}\text{\,}{\mathrm{s}}^{-2}$