"Motion Control for Mobile Robot Navigation Using Machine Learning: a Survey"
-
[
2020
] [📝] [ 🎓University of Texas
] [ 🚗US Army Research Laboratory
] -
[
survey
,classical planning
,learning-based planning
,hybrid
,parameter tuning
,engineering efforts
]
Click to expand
Within the navigation stack, learning-based approaches can replace (1 ) the entire navigation stack (end-to-end ), (2 ) navigation subsystems (either global or local ), and (3 ) individual components (e.g. planner parameters, world representation). The authors advocate to use ML at the subsystem or component level. And to forget about end-to-end : the red cross shows that no end-to-end approach outperforms classical systems. Last but not least: ML looks promising for social navigation and reactive local planning. Source |
Authors: Xiao, X., Liu, B., Warnell, G., & Stone, P.
-
Motivation:
- Compare between the
classical
planners and emerginglearning-based
paradigms forplanning
(navigation
andmotion control
).
- Compare between the
-
Some interesting conclusions:
-
Scope:
-
Most learning-based works focus on
end-to-end
approaches.- Because appealing aspects, e.g. they should be able to avoid cascading errors such as errors in the
global
representation being propagated to thelocal
planning level. - Nevertheless limitations, e.g. they lack of proven reliable applicability in real-world scenarios.
- Because appealing aspects, e.g. they should be able to avoid cascading errors such as errors in the
-
When focusing on subsystems, most address the
local
rather than theglobal
planning.- Because it is easier to generalize to a close
local
goal orlocal
segment from a global path than an unseen, farawayglobal
goal.
- Because it is easier to generalize to a close
-
-
Performance:
-
Very few
learning
-based approaches actually improve uponclassical
techniques.- More "proofs of concepts": the research community initially focused on answering the question of whether or not navigation was even possible with learning approaches.
-
Safety
is overlooked byML
.-
"Only
11
out of the72
surveyed papers describe navigation systems that can provide some form of safety assurance, and each of these does so using a component from aclassical
system."
-
-
end-to-end
: one third of the approaches surveyed lacks the ability to navigate to user-defined goals.-
"It is very difficult for a learned model to generalize to arbitrarily defined goals, which is not seen in its training data. Not being able to see similar goal configurations during training makes it hard to generalize to arbitrarily specified goals."
- All of the non-
end-to-end
approaches surveyed can do that.
-
-
end-to-end
:Explainability
is overlooked.-
"None of the
43
end-to-end
approaches surveyed even attempted to maintain any notion of explainability."
-
-
-
Improvements:
-
Can learning approaches enable new navigation capabilities that are difficult to achieve with classical techniques?
- Yes, e.g. social navigation.
-
Are learning approaches really getting rid of the extensive human engineering (e.g., filtering, designing, modelling, tuning, etc.)? E.g. for manual parameter tuning.
- Not so clear!
- The "learning burden" (high demand on training data, hyperparameter search) cannot always be automated and is insufficiently acknowledged.
-
"We found that the literature on learning-based navigation did little to explicitly acknowledge or characterize the very real costs of hyperparameter search and training overhead."
-
-
Main recommendation: Combine the two paradigms.
-
Forget about
end-to-end
.-
"The current best practice is to use
ML
at the subsystem or component level." - In other words: do not replace a "classical" component that already works. Rather:
- Replace small "classical" components which suffer from limitations: e.g. parameter tuning, social interaction modelling.
- Extend them: e.g. they are currently not able to continuously adapt their parameters from experiences collected during deployment.
-
-
Example:
- Learning modules to adapt to new environments.
- E.g. dynamically changing parameters.
- Classical components to ensure that the system obeys hard
safety
constraints.
- Learning modules to adapt to new environments.
-
Example:
ML
to tune the parameters of a classical system. E.g. by trial and error.
-
Example:
- Continuous (lifelong) learning:
ML
to improve based on real deployment experience- Risk: catastrophic forgetting.
- Continuous (lifelong) learning:
-
Example:
ML
for the reactivelocal
planning level.- These reactive behaviours are difficult to model using rule-based symbolic reasoning, but are ripe for learning from experience.
- Analogy to human navigation:
-
"Humans typically perform high-level deliberation to come up with long-range plans, such as how to get from a house or apartment to a local park, but navigation becomes much more reactive at the local level when needing to respond to immediate situations, such as avoiding a running dog or moving through extremely complex or tight spaces."
-
- Especially: social behaviours.
-
"An Autonomous Driving Framework for Long-term Decision-making and Short-term Trajectory Planning on Frenet Space"
-
[
2020
] [📝] [] [ 🎓University of California
] -
[
behavioural planner
,hierarchy
,safety supervisor
,IDM
,MOBIL
,Frenet frame
,lattices
]
Click to expand
Hierarchical structure: the BP decides to change lane or not based on a state machine. This short term goal, together with a desired speed , are processed by the local planner (LP ) which generates a trajectory . This trajectory is tracked by a low-level controller using two PID s. On the top, a supervisor runs at higher frequency. It can quickly react and overwrite the goals set by BP . Source |
The driving style can be adjusted with 2 parameters: the desired gap with leader when in StayOnLane and the desired gap with follower when ChangeLane . Source |
Authors: Moghadam, M., & Elkaim, G. H.
-
Motivation:
1-
Modularity and hierarchy.- By nature, modules run at different frequencies.
2-
Enable different driving styles.- Here {
agile
,moderate
,conservative
} styles with different values for two parameters: 1-
Politeness
factor inMOBIL
which influence tailgating, i.e. gap with the following car.2-
Safe time headway
inIDM
, i.e. gap with the leading car.
- Here {
- Task: Highway driving.
-
Four levels (from
bottom
totop
): -
1-
Trajectory following
+Drive
andsteer
control.- Task: stabilize the vehicle dynamics while tracking the commanded
trajectory
. - Input: a spatiotemporal
trajectory
:τ
= { (s
(t
),d
(t
)) ∈R²
t
∈ [0
,T
] }. - Output:
steering
anddrive
commands. - How:
- Points (
x
,y
,speed
) are sampled from thetrajectory
. - Two
PID
controllers are then used to track the desiredspeed
andWP
s. - For
lateral
control: Atwo-point
control model.
- Points (
- Task: stabilize the vehicle dynamics while tracking the commanded
-
2-
Trajectory planning
. Also called "Local Planner" (LP
).-
Task: produce a safe, feasible and optimal
trajectory
to meet the defined short-term goal.-
"The planner translates the commanded
maneuver modes
, such asLaneChangeRight
andStayOnTheLane
, along with thedesired speed
to optimal trajectories in a variable time-horizon window."
-
-
Input: a
target speed
and amanoeuvre mode
in {LaneChangeRight
,LaneChangeLeft
,StayOnTheLane
}. -
Output: spatiotemporal
trajectory
.- It consists of two polynomials of time, concerning
lateral
andlongitudinal
movements.
- It consists of two polynomials of time, concerning
-
Frenet
frame.-
"Transferring the calculations to
Frenet
space makes the driving behavior invariant to the road curvatures and road slopes in3
dimensions, which improves the optimization computations significantly and simplifies thecost
function manipulation."
-
-
Optimization: generate
lattices
+ evaluate + pick the best one.-
"Six constraints for
d(t)
and five fors(t)
, which makes them form quintic (r = 5
) and quartic (k = 4
) polynomials, respectively." -
[
3
degrees of freedom] "Sincet0
andTi−1
are known values at each time-step, producing lattices boils down to identifying terminal manifolds: arrival timetf
, lateral positiondf
, and speedvf
. The setT
is produced by varying these3
unknowns within the feasible ranges." - Surrounding cars are assumed to maintain their speeds.
-
[After filtering based on hard constraints] "The remaining trajectory candidates are examined in terms of
velocity
andacceleration
that aresafe
, dynamicallyfeasible
, andlegally
allowed."
-
-
What if
BP
decisions are found to be non-applicable?LP
can overwriteBP
.-
"Similar to target
lane
, thespeed
commanded by theIDM
can also be overwritten by theLP
before being sent to low- level controller."
-
-
3-
Behavioural planner (BP
).- Task: decide high-level short-term goals.
- Input: scene description. Probably a list of <
pos
,speed
,heading
>. - Output: a
target speed
and amanoeuvre mode
. IDM
andMOBIL
models used for:- The
transitions
of the state machine. - The computation of the
target speed
. -
"The ego vehicle stays in the current lane with a desired speed computed by
IDM
module untilMOBIL
algorithm decides on a lane change."
- The
-
4-
Safety
Supervisor.- Task: monitor decisions and react quickly to
risky
situations.-
"The supervisor sends a recalculation command to the modules if an unpredicted situation appears during the path following."
-
- Frequency: higher that
BP
andLP
.-
[Need to be able to react] "The surrounding human/autonomous drivers may perform unexpected maneuvers that jeopardize the safety of the generated trajectory."
-
- To avoid forward collisions: use a second
IDM
with more conservative parameters thanBP
.-
"If
TTC
violates the safety threshold,IDM2
raises the safety violation flag, and the supervisor calls theLP
to recalculate the trajectory based on the current situation."
-
- Task: monitor decisions and react quickly to
"Formalizing Traffic Rules for Machine Interpretability"
-
[
2020
] [📝] [ 🚗Fortiss
] -
[
LTL
,Vienna convention
,SPOT
,INTERACTION Dataset
]
Click to expand
Top left: Different techniques on how to model the rules have been employed: formal logics such as Linear Temporal Logic LTL or Signal Temporal Logic (STL ), as well as real-value constraints. Middle and bottom: Rules are separated into premise and conclusion . The initial premise and exceptions (red) are combined by conjunction . Source |
Authors: Esterle, K., Gressenbuch, L., & Knoll, A.
-
Motivation:
-
"Traffic rules are fuzzy and not well defined, making them incomprehensible to machines."
- The authors formalize traffic rules from legal texts (here
StVO
) to a formal language (hereLTL
).
-
-
Which legal text defines rules?
- For instance the Straßenverkehrsordnung (
StVO
), which is the German concretization of theVienna Convention
on Road Traffic.
- For instance the Straßenverkehrsordnung (
-
Why Linear Temporal Logic (
LTL
) as the formal language to specify traffic rules?-
"During the legal analysis,
conjunction
,disjunction
,negation
andimplication
proved to be powerful and useful tools for formalizing rules. As traffic rules such asovertaking
consider temporal behaviors, we decided to useLTL
." -
"Others have used Signal Temporal Logic (
STL
) to obtain quantitative semantics about rule satisfaction. Quantitive semantics might be beneficial for relaxing the requirements to satisfy a rule."
-
-
Rules are separated into
premise
andconclusion
.-
"This allows rules to be separated into a
premise
about the current state of the environment, i.e. when a rule applies, and the legal behavior of the ego agent in that situation (conclusion
). Then, exceptions to the rules can be modeled to be part of the assumption."
-
-
Tools:
INTERACTION
: a dataset which focuses on dense interactions and analyze the compliance* of each vehicle to the traffic rules.SPOT
: aC++
library for model checking, to translate the formalizedLTL
formula to a deterministic finite automaton and to manipulate the automatons.BARK
: a benchmarking framework.
-
Evaluation of rule-violation on public data:
-
"Roughly every fourth
lane change
does not keep asafe distance
to the rear vehicle, which is similar for theGerman
andChinese
Data."
-
"A hierarchical control system for autonomous driving towards urban challenges"
-
[
2020
] [📝] [ 🎓Chungbuk National University, Korea
] -
[
FSM
]
Click to expand
Behavioural planning is performed using a two-state FSM . Right: transition conditions in the M-FSM . Source |
Authors: Van, N. D., Sualeh, M., Kim, D., & Kim, G. W.
- Motivations:
-
"In the
DARPA
Urban Challenge, StanfordJunior
team succeeded in applyingFSM
with several scenarios in urban traffic roads. However, the main drawback ofFSM
is the difficulty in solving uncertainty and in large-scale scenarios." - Here:
- The uncertainty is not addressed.
- The diversity of scenarios is handled by a two-stage Finite State Machine (
FSM
).
-
- About the two-state
FSM
:1-
A MissionFSM
(M-FSM
).- Five states:
Ready
,Stop-and-Go
(SAG
) (main mode),Change-Lane
(CL
),Emergency-stop
,avoid obstacle mode
.
- Five states:
2-
A ControlFSM
(C-FSM
) in eachM-FSM
state.
- The
decision
is then converted intospeed
andwaypoints
objectives, handled by the local path planning.- It uses a real-time hybrid
A*
algorithm with an occupancy grid map. - The communication
decision
->path planner
is unidirectional: No feedback is given regarding the feasibility for instance.
- It uses a real-time hybrid
"Trajectory Optimization and Situational Analysis Framework for Autonomous Overtaking with Visibility Maximization"
- [
2019
] [📝] [🎞️] [ 🎓National University of Singapore
,Delft University
,MIT
] - [
FSM
,occlusion
,partial observability
]
Click to expand
Left: previous work Source. Right: The BP FSM consists in 5 states and 11 transitions. Each transition from one state to the other is triggered by specific alphabet unique to the state. For instance, 1 is Obstacle to be overtaken in ego lane detected . Together with the MPC set of parameters, a guidance path is passed to the trajectory optimizer. Source. |
Authors: Andersen, H., Alonso-mora, J., Eng, Y. H., Rus, D., & Ang Jr, M. H.
- Main motivation:
- Deal with occlusions, i.e. partial observability.
- Use case: a car is illegally parked on the vehicle’s ego lane. It may fully occlude the visibility. But has to be overtaken.
- One related works:
- "Trajectory Optimization for Autonomous Overtaking with Visibility Maximization" - (Andersen et al., 2017)
- [🎞️].
- [🎞️].
- [🎞️].
- About the hierarchical structure.
1-
A high-level behaviour planner (BP
).- It is structured as a deterministic finite state machine (
FSM
). - States include:
Follow ego-lane
Visibility Maximization
Overtake
Merge back
Wait
- Transition are based on some deterministic
risk assessment
.- The authors argue that the deterministic methods (e.g. formal verification of trajectory using
reachability analysis
) are simpler and computationally more efficient than probabilistic versions, while being very adapted for this information maximization: -
This is due to the fact that the designed behaviour planner explicitly breaks the traffic rule in order to progress along the vehicle’s course.
- The authors argue that the deterministic methods (e.g. formal verification of trajectory using
- It is structured as a deterministic finite state machine (
- Interface
1-
>2-
:- Each state correspond to specific set of parameters that is used in the trajectory optimizer.
-
"In case of
Overtake
, a suggested guidance path is given to both theMPC
and `backup trajectory generator".
2-
A trajectory optimizer.- The problem is formulated as receding horizon planner and the task is to solve, in real-time, the non-linear constrained optimization.
- Cost include
guidance path deviation
,progress
,speed deviation
,size of blind spot
(visible area) andcontrol inputs
. - Constraints include, among other,
obstacle avoidance
. - The prediction horizon of this
MPC
is5s
.
- Cost include
- Again (I really like this idea),
MPC
parameters are set by theBP
.- For instance, the cost for
path deviation
is high forFollow ego-lane
, while it can be reduced forVisibility Maximization
. -
"Increasing the visibility maximization cost resulted in the vehicle deviating from the path earlier and more abrupt, leading to frequent wait or merge back cases when an oncoming car comes into the vehicle’s sensor range. Reducing visibility maximization resulted in later and less abrupt deviation, leading to overtaking trajectories that are too late to be aborted. We tune the costs for a good trade-off in performance."
- Hence, depending on the state, the task might be to maximize the amount of information that the autonomous vehicle gains along its trajectory.
- For instance, the cost for
- The problem is formulated as receding horizon planner and the task is to solve, in real-time, the non-linear constrained optimization.
-
"Our method considers visibility as a part of both
decision-making
andtrajectory generation
".
"Jointly Learnable Behavior and Trajectory Planning for Self-Driving Vehicles"
- [
2019
] [📝] [ 🚗Uber
] - [
max-margin
]
Click to expand
Source. |
Authors: Sadat, A., Ren, M., Pokrovsky, A., Lin, Y., Yumer, E., & Urtasun, R.
- Main motivation:
- Design a decision module where both the behavioural planner and the trajectory optimizer share the same objective (i.e. cost function).
- Therefore "joint".
-
"[In approaches not-joint approaches] the final trajectory outputted by the trajectory planner might differ significantly from the one generated by the behavior planner, as they do not share the same objective".
- Requirements:
1-
Avoid time-consuming, error-prone, and iterative hand-tuning of cost parameters.- E.g. Learning-based approaches (
BC
).
- E.g. Learning-based approaches (
2-
Offer interpretability about the costs jointly imposed on these modules.- E.g. Traditional modular
2
-stage approaches.
- E.g. Traditional modular
- About the structure:
- The driving scene is described in
W
(desired route
,ego-state
,map
, anddetected objects
). ProbablyW
for "World"? - The behavioural planner (
BP
) decides two things based onW
:1-
An high-level behaviourb
.- The path to converge to based on one chosen manoeuvre:
keep-lane
,left-lane-change
, orright-lane-change
. - The
left
andright
lane boundaries. - The obstacle
side assignment
: whether an obstacle should stay in thefront
,back
,left
, orright
to the ego-car.
- The path to converge to based on one chosen manoeuvre:
2-
A coarse-level trajectoryτ
.- The loss has also a regularization term.
- This decision is "simply" the
argmin
of the shared cost-function, obtained by sampling+selecting the best.
- The "trajectory optimizer" refines
τ
based on the constraints imposed byb
.- For instance an overlap cost will be incurred if the
side assignment
ofb
is violated.
- For instance an overlap cost will be incurred if the
- A cost function parametrized by
w
assesses the quality of the selected <b
,τ
> pair:cost
=w^T
.sub-costs-vec
(τ
,b
,W
).- Sub-costs relate to safety, comfort, feasibility, mission completion, and traffic rules.
- The driving scene is described in
- Why "learnable"?
- Because the weight vector
w
that captures the importance of each sub-cost is learnt based on human demonstrations.-
"Our planner can be trained jointly end-to-end without requiring manual tuning of the costs functions".
-
- They are two losses for that objective:
1-
Imitation loss (withMSE
).- It applies on the <
b
,τ
> produced by theBP
.
- It applies on the <
2-
Max-margin loss to penalize trajectories that have small cost and are different from the human driving trajectory.- It applies on the <
τ
> produced by the trajectory optimizer. -
"This encourages the human driving trajectory to have smaller cost than other trajectories".
- It reminds me the
max-margin
method inIRL
where the weights of the reward function should make the expert demonstration better than any other policy candidate.
- It applies on the <
- Because the weight vector
"Liability, Ethics, and Culture-Aware Behavior Specification using Rulebooks"
-
[
2019
] [📝] [] [🎞️] [🎞️] [ 🎓ETH Zurich
] [ 🚗nuTonomy
,Aptiv
] -
[
sampling-based planning
,safety validation
,reward function
,RSS
]
Click to expand
Some figures:
Defining the rulebook . Source. |
The rulebook is associated to an operator =< to prioritize between rules. Source. |
The rulebook serves for deciding which trajectory to take and can be adapted using a series of operations. Source. |
Authors: Censi, A., Slutsky, K., Wongpiromsarn, T., Yershov, D., Pendleton, S., Fu, J., & Frazzoli, E.
-
Allegedly how nuTonomy (an Aptiv company) cars work.
-
One main concept: "rulebook".
- It contains multiple
rules
, that specify the desired behaviour of the self-driving cars. - A rule is simply a scoring function, or “violation metric”, on the realizations (= trajectories).
- The degree of violation acts like some penalty term: here some examples of evaluation of a realization
x
evaluated by a ruler
:- For speed limit:
r
(x
) = interval for which the car was above45 km/h
. - For minimizing harm:
r
(x
) = kinetic energy transferred to human bodies.
- For speed limit:
- Based on Use as a comparison operator to rank candidate trajectories.
- It contains multiple
-
One idea: Hierarchy of rules.
- With many rules being defined, it may be impossible to find a realization (e.g. trajectory) that satisfies all.
- But even in critical situation, the algorithm must make a choice - the least catastrophic option (hence no concept of infeasibility.)
- To deal with this concept of "Unfeasibility", priorities between conflicting rules which are therefore hierarchically ordered.
- Hence a rulebook
R
comes with some operator<
: <R
,<
>. - This leads to some concepts:
- Safety vs. infractions.
- Ex.: a rule "not to collide with other objects" will have a higher priority than the rule "not crossing the double line".
- Liability-aware specification.
- Ex.: (edge-case): Instruct the agent to collide with the object on its lane, rather than collide with the object on the opposite lane, since changing lane will provoke an accident for which it would be at fault.
- This is close to the RSS ("responsibility-sensitive safety" model) of Mobileye.
- Hierarchy between rules:
- Top: Guarantee safety of humans.
- This is written analytically (e.g. a precise expression for the kinetic energy to minimize harm to people).
- Bottom: Comfort constraints and progress goals.
- Can be learnt based on observed behaviour (and also tend to be platform- and implementation- specific).
- Middle: All the other priorities among rule groups
- There are somehow open for discussion.
- Top: Guarantee safety of humans.
-
How to build a rulebook:
- Rules can be defined analytically (e.g.
LTL
formalism) or learnt from data (for non-safety-critical rules). - Violation functions can be learned from data (e.g.
IRL
). - Priorities between rules can also be learnt.
- Rules can be defined analytically (e.g.
-
One idea: manipulation of rulebooks.
- Regulations and cultures differ depending on the country and the state.
- A rulebook <
R
,<
> can easily be adapted using three operations (priority refinement
,rule augmentation
,rule aggregation
).
-
Related work: Several topics raised in this paper reminds me subjects addressed in Emilio Frazzoli, CTO, nuTonomy - 09.03.2018
- 1- Decision making with FSM:
- Too complex to code. Easy to make mistake. Difficult to adjust. Impossible to debug (:cry:).
- 2- Decision making with E2E learning:
- Appealing since there are too many possible scenarios.
- But how to prove that and justify it to the authorities?
- One solution is to revert such imitation strategy: start by defining the rules.
- 3- Decision making "cost-function-based" methods
- 3-1-
RL
/MCTS
: not addressed here. - 3-2- Rule-based (not the
if
-else
-then
logic but rather traffic/behaviour rules).
- 3-1-
- First note:
- Number of rules: small (
15
are enough for level-4
). - Number of possible scenarios: huge (combinational).
- Number of rules: small (
- Second note:
- Driving baheviours are hard to code.
- Driving baheviours are hard to learn.
- But driving baheviours are easy to assess.
- Strategy:
- 1- Generate candidate trajectories
- Not only in time and space.
- Also in term of semantic (logical trajectories in Kripke structure).
- 2- Check if they satisfy the constraints and pick the best.
- This involves linear operations.
- 1- Generate candidate trajectories
- Conclusion:
-
"Rules and rules priorities, especially those that concern safety and liability, must be part of nation-wide regulations to be developed after an informed public discourse; it should not be up to engineers to choose these important aspects."
- This reminds me the discussion about social-acceptance I had at IV19.^
- As E. Frazzoli concluded during his talk, the remaining question is:
- "We do not know how we want human-driven vehicle to behave?"
- Once we have the answer, that is easy.
-
- 1- Decision making with FSM:
Some figures from this related presentation:
Candidate trajectories are not just spatio-temporal but also semantic. Source. |
Define priorities between rules, as Asimov did for his laws. Source. |
As raised here by the main author of the paper, I am still wondering how the presented framework deals with the different sources of uncertainties. Source. |
"Provably Safe and Smooth Lane Changes in Mixed Traffic"
Click to expand
Some figures:
The first safe? check might lead to conservative behaviours (huge gaps would be needed for safe lane changes). Hence it is relaxed with some Probably Safe? condition. Source. |
Source. |
Formulation by Pek, Zahn, & Althoff, 2017. Source. |
Authors: Naumann, M., Königshof, H., & Stiller, C.
-
Main ideas:
- The notion of safety is based on the responsibility sensitive safety (
RSS
) definition.- As stated by the authors, "A
safe
lane change is guaranteed not tocause
a collision according to the previously defined rules, while a single vehicle cannot ensure that it will never be involved in a collision."
- As stated by the authors, "A
- Use set-based reachability analysis to prove the "RSS-safety" of lane change manoeuvre based on gap evaluation.
- In other words, it is the responsibility of the ego vehicle to maintain safe distances during the lane change manoeuvre.
- The notion of safety is based on the responsibility sensitive safety (
-
Related works: A couple of safe distances are defined, building on
RSS
principles (after IV19, I tried to summarize some of the RSS concepts here).- "Verifying the Safety of Lane Change Maneuvers of Self-driving Vehicles Based on Formalized Traffic Rules", (Pek, Zahn, & Althoff, 2017)
"Decision-Making Framework for Autonomous Driving at Road Intersections: Safeguarding Against Collision, Overly Conservative Behavior, and Violation Vehicles"
-
[
2018
] [📝] [🎞️] [ 🎓Daejeon Research Institute, South Korea
] -
[
probabilistic risk assessment
,rule-based probabilistic decision making
]
Click to expand
One figure:
Source. |
Author: Noh, S.
- Many ML-based works criticize rule-based approaches (over-conservative, no generalization capability and painful parameter tuning).
- True, the presented framework contains many parameters whose tuning may be tedious.
- But this approach just works! At least they go out of the simulator and show some experiments on a real car.
- I really like their video, especially the multiple camera views together with the
RViz
representation. - It can be seen that probabilistic reasoning and uncertainty-aware decision making are essential for robustness.
- One term: "Time-to-Enter" (tte).
- It represents the time it takes a relevant vehicle to reach the potential collision area (CA), from its current position at its current speed.
- To deal with uncertainty in the measurements, a variant of this heuristic is coupled with a Bayesian network for probabilistic threat-assessment.
- One Q&A: What is the difference between situation awareness and situation assessment?
- In situation awareness, all possible routes are considered for the detected vehicles using a map. The vehicles whose potential route intersect with the ego-path are classified as relevant vehicles.
- In situation assessment, a threat level in {
Dangerous
,Attentive
,Safe
} is inferred for each relevant vehicle.
- One quote:
"The existing literature mostly focuses on motion prediction, threat assessment, or decision-making problems, but not all three in combination."