Zenseact Ab

Zenuity Ab

Magnus Gyllenhammar

Majid Khorsand Vakilzadeh

Carl Zandén

Håkan Sivencrona

Mina Alibeigi

Andreas Valencia Falkovén

Adam Tonderski

Andreas Falkovén

Joakim Olsson

Kilian Zwirglmaier

Cagdas Ulas

The present disclosure relates to a method performed by a buffer resources prioritizing system for storage prioritization in an event buffer configured to continuously collect operational data of an Automated Driving System, ADS, of a vehicle. The buffer resources prioritizing system obtains sensor data of one or more sensors onboard the vehicle. The buffer resources prioritizing system further determines, at least partly based on the sensor data, current ADS-related operational conditions at least comprising states of vehicle surroundings and internal states of the vehicle. Moreover, the buffer resources prioritizing system determines an upcoming scene predicted to evolve from the current operational conditions. Furthermore, the buffer resources prioritizing system deduces based on assessment of the predicted scene, a storage priority score thereof reflecting predicted relevance of freezing event data of the predicted scene in the event buffer. The disclosure also relates to a buffer resources prioritizing system.

Event buffer storage prioritization

A path planning method and system for a vehicle. The method includes obtaining risk map of a surrounding environment of vehicle. Risk map is formed based on an actuation capability of the vehicle and location of free-space areas in the surrounding environment, actuation capability including uncertainty estimation for actuation capability and the location of free-space areas comprising an uncertainty estimation for the estimated location of free-space areas. Risk map includes risk parameter for each of a plurality of area segments included in the surrounding environment of the vehicle. Obtaining at least one candidate path for vehicle, determining total risk value for each candidate path based on risk parameters of a set of area segments intersected by the at least one path, selecting a candidate path, of at least one candidate path, fulfilling a risk value criterion, and generating, at an output, a first signal indicative of selected candidate path.

Path planning in autonomous driving environments

A control system and a method for estimating a risk exposure of an automated driving system (ADS) of a vehicle. Obtaining an actuation capability of the vehicle, wherein the obtained actuation capability includes an uncertainty estimation for the actuation capability. Obtaining a location of free-space areas in the surrounding environment of the vehicle, wherein the obtained location of free-space areas comprises an uncertainty estimation for the estimated location of free-space areas. Forming a risk map of the surrounding environment of the vehicle based on the obtained actuation capability and the obtained location of free-space areas, wherein the risk map includes a risk parameter for each of a plurality of area segments included in the surrounding environment of the vehicle. Determining a total risk value of the ADS based on the risk parameters of a set of area segments intersected by at least one planned path of the ADS.

Risk estimation in autonomous driving environments

The present invention relates to methods and systems that utilize the production vehicles to develop new perception features related to new sensor hardware as well as new algorithms for existing sensors by using federated learning. To achieve this, the production vehicle's own worldview is post-processed and used as a reference, towards which the output of the software (SW) or hardware (HW) under development is compared. Through this comparison, a cost function can be calculated and an update of the SW parameters can be locally updated according to this cost function.

Platform for perception system development for automated driving systems

The present disclosure describes a method for monitoring operations of an automated driving system (ADS) of a vehicle. For each monitored operation the method includes: determining a geographical position of the vehicle; determining an intended path of the vehicle; and determining one or more intended parameters associated with performing a driving manoeuvre of said vehicle from the determined geographical position along the intended path. For each monitored operation the method further includes: obtaining one or more parameters associated with performing the driving manoeuvre of said vehicle from said determined geographical position; and retrieving, from a statistical model, data indicative of a statistical distribution related to one or more corresponding intended and/or obtained parameters for said intended path. Based on said retrieved data, determining whether there is an anomaly associated with said monitored operation; and taking at least one action of a set of predefined actions if an anomaly is determined.

Method and device for monitoring operations of an automated driving system of a vehicle

The present invention relates to a method and apparatus that utilize production vehicles to develop new path planning features for Automated Driving Systems (ADSs) by using federated learning. To achieve this the 'under-test' path planning module's output is evaluated in closed-loop in order to produce a cost-function that is subsequently used to update or train a path planning model of the path planning development-module.

Platform for path planning system development for automated driving system

The present invention relates to a method and apparatus that utilize production vehicles to develop new path planning features for Automated Driving Systems (ADSs) by using federated learning. To achieve this the 'under-test' path planning module's output is evaluated in open-loop in order to produce a cost-function that is subsequently used to update or train a path planning model of the path planning module.

A computer implemented method and related aspects for updating a perception function of a plurality of vehicles having an Automated Driving System (ADS) are disclosed. The method includes obtaining one or more locally updated model parameters of a self-supervised machine-learning algorithm from a plurality of remote vehicles, and updating one or more model parameters of a global self-supervised machine-learning algorithm based on the obtained one or more locally updated model parameters. Further, the method includes fine-tuning the global self-supervised machine-learning algorithm based on an annotated dataset in order to generate a fine-tuned global machine-learning algorithm comprising one or more fine-tuned model parameters. The method further includes forming a machine-learning algorithm for an in-vehicle perception module based on the fine-tuned global machine-learning algorithm, and transmitting one or more model parameters of the formed machine-learning algorithm for the in-vehicle perception module to the plurality of remote vehicles.

System and method for federated learning of self-supervised networks in automated driving systems

A test-optimizing system for supporting shadow mode testing of ADS software included in an ADS-provided vehicle. One or more potential vehicle routes are retrieved. A geographical area covering the one or more potential vehicle routes is obtained and data of crucial locations associated with past vehicle situations identified as critical and/or challenging. Moreover, the test-optimizing system retrieves respective ODD for one or more candidate software respectively adapted to run in the background of the vehicle. ODD-compliant locations for respective candidate software are determined, by identifying locations out of the data of crucial locations lying within respective candidate software's ODD. At least a first test-compliant location along at least a first route out of the one or more potential vehicle routes is determined by identifying for at least a first candidate software, locations out of the ODD-compliant locations situated along the at least first route.

Vehicle software shadow mode testing

A method for identifying scenarios of interest for development, verification and/or validation of an ADS of vehicle. Obtaining risk map of surrounding environment of vehicle, risk map is formed based on actuation capability of vehicle and location of free-space areas in surrounding environment. The actuation capability comprises uncertainty estimation for actuation capability and location of free-space areas comprises uncertainty estimation for estimated location of free-space areas. Risk map includes risk parameter for each of a plurality of area segments comprised in surrounding environment of vehicle. Determining compounded risk value of ADS based on risk parameters of a set of area segments of risk map. Monitoring scenario trigger by monitoring at least one of determined compounded risk value against compounded risk trigger threshold, a development of risk map over time against a map volatility trigger threshold, and a development of compounded risk value over time against a risk volatility threshold.

Scenario identification in autonomous driving environments

The present invention relates to methods and systems that utilize the production vehicles to develop new perception features related to new sensor hardware as well as new algorithms for existing sensors by using federated learning. To achieve this, the production vehicle's own worldview is post-processed and used as a reference, towards which the output of the software (SW) or hardware (HW) under development is compared. In case of a large discrepancy between the baseline worldview and perceived worldview by the module-under-test, the data is weakly annotated by the baseline worldview. Such weakly annotated data may subsequently be used to update the SW parameters of the 'perception model' in the module-under-test in each individual vehicle, or to be transmitted to the “back-office” for off-board processing or more accurate annotations.

Platform for perception system development for automated driving system

An assessment system for performance evaluation and updating of a PMUD of an ego-vehicle. The assessment system obtains world view data from a perception module configured to generate the world view data based on sensor data obtained from vehicle-mounted sensors; obtains other world view data generated by another perception module; forms a joint world view by matching the world view data the other world view data; and obtains perception data based on a perception model and sensor data obtained from one or more vehicle-mounted sensors. The assessment system further matches the perception data to the formed joint world view; evaluates the obtained perception data in reference to the joint world view to determine an estimation deviation in an identified match between the perceptive parameter of the perception data and a corresponding perceptive parameter in the joint world view; and updates parameters of the perception model based on the estimation deviation.

ADS perception development

An apparatus and method performed by a perception comparing system of a vehicle for perception performance evaluation of an ADAS or ADS. The perception comparing system establishes communication with a secondary vehicle determined and/or estimated to be positioned within a potential range of surrounding detecting sensors on-board the vehicle. The system derives perception data from a perception system of the ADAS or ADS. The system receives secondary perception data from a secondary perception system of a secondary ADAS or ADS of the secondary vehicle and determines a discrepancy output based on comparison of at least a portion of the both of perception data and the secondary perception data. The system communicates acknowledgement data when at least a portion of the discrepancy output fulfils discrepancy exceedance criteria and/or when the vehicle is perceivable in the secondary perception data but the secondary vehicle not is locatable in the perception data.

Growing community of inventors

Pixbo, Sweden

Magnus Gyllenhammar