Collision prevention system and method

Abstract

We generally describe a collision prevention system (100) comprising: a localization system (402) for determining positions of an autonomous vehicle (104) and a human (106); and a collision determination unit (404) coupled to or in communication with the localization system (402), wherein the collision determination unit (404) is configured to determine, based on the determined positions of the autonomous vehicle (104) and the human (106), whether a predefined condition for an anticipated collision of the autonomous vehicle (104) with the human (106) is met; wherein the collision prevention system (100) is configured to: lock the autonomous vehicle (104) if the predefined condition is met; alert the human (106) for whom the predefined condition for colliding with the autonomous vehicle (104) is met; and allow unlocking of the autonomous vehicle (104) to be performed or initialized by the alerted human (106) only.

Claims

1. A collision prevention system, comprising: a localization system configured to determine positions of an autonomous vehicle and a human; and a collision determination unit coupled to or in communication with the localization system, wherein the collision determination unit is configured to determine, based on the determined positions of the autonomous vehicle and the human, whether a predefined condition for an anticipated collision of the autonomous vehicle with the human is met; wherein the collision prevention system is configured to: lock the autonomous vehicle if the predefined condition is met; alert the human for whom the predefined condition for colliding with the autonomous vehicle is met; and allow unlocking of the autonomous vehicle to be performed or initialized by the alerted human only.

2. The collision prevention system of claim 1, wherein the collision prevention system is configured to determine a predicted path of movement of one or both of the autonomous vehicle and the human to determine whether the predefined condition is met.

3. The collision prevention system of claim 1, wherein the collision determination unit is configured to determine whether the predefined condition is met based on position streams obtained via the localization system.

4. The collision prevention system of claim 1, wherein localization system is configured to determine the positions of the autonomous vehicle and the human by determining relative positions between the autonomous vehicle and the human.

5. The collision prevention system of claim 1, wherein the localization system comprises one or more tags fixable to the autonomous vehicle and the human for determining: absolute positions of the autonomous vehicle and/or the human; and/or relative positions between the autonomous vehicle and the human.

6. The collision prevention system of claim 1, wherein the localization system comprises one or more anchors configured to perform positioning measurements of the autonomous vehicle and the human.

7. The collision prevention system of claim 6: wherein the localization system comprises one or more tags fixable to the autonomous vehicle and the human for determining: absolute positions of the autonomous vehicle and/or the human; and/or relative positions between the autonomous vehicle and the human; wherein the one or more anchors are in communication with the one or more tags for determining the position of the autonomous vehicle and/or the human.

8. The collision prevention system of claim 6, wherein the one or more anchors are mobile anchors configured to determine a relative position between the autonomous vehicle and the human.

9. The collision prevention system of claim 1, wherein the collision determination unit is configured to determine whether the predefined condition is met based on one or more maps, each of the maps depicting one or more time-varying zones in which one or both of the autonomous vehicle and the human can move.

10. The collision prevention system of claim 1, wherein the collision determination unit is configured to determine whether the predefined condition is met based on historical data relating to: collisions of autonomous vehicles with humans; and/or determined positions or trajectories of the autonomous vehicles and humans.

11. The collision prevention system of claim 10, wherein the collision determination unit is configured to determine whether the predefined condition is met based on a machine learning algorithm developed using the historical data.

12. The collision prevention system of claim 1, wherein the collision prevention system is further configured to define a first danger zone around the autonomous vehicle, or around a trajectory of the autonomous vehicle, within which first danger zone it is possible for the predefined condition to be met.

13. The collision prevention system of claim 12, wherein the first danger zone is defined taking into consideration: an uncertainty in position determination of one or both of the autonomous vehicle and the human; a reaction time of a closed loop vehicle control; a prediction of the trajectory of the autonomous vehicle; and/or one or more movement characteristics of the human.

14. The collision prevention system of claim 12, wherein the collision prevention system is further configured to alert all humans located within the first danger zone.

15. The collision prevention system of claim 14, wherein the collision prevention system is further configured to allow unlocking of the autonomous vehicle only if it is triggered by each of the humans located within the first danger zone.

16. The collision prevention system of claim 1, wherein the collision prevention system is configured to allow unlocking of the autonomous vehicle to be triggered by: pressing a physical button and/or reading an ID unique to the alerted human; a mobile device of the alerted human sending a safety signal to the collision determination unit; and/or detection of the alerted human leaving a zone which is within reach by the autonomous vehicle or by a predicted trajectory of the autonomous vehicle.

17. The collision prevention system of claim 1, wherein the collision prevention system is configured to define a second danger zone and to allow unlocking of the autonomous vehicle to be triggered only: if no humans are located within the second danger zone; and/or by a signal sent from outside the second danger zone.

18. A method, comprising: determining, based on real-time locations of a human and an autonomous vehicle, whether a predefined condition for an anticipated collision of the autonomous vehicle with the human is met; locking the autonomous vehicle if the predefined condition is met; alerting the human for whom the predefined condition for colliding with the autonomous vehicle is met; wherein a subsequent unlocking of the autonomous vehicle is performable or initializable by the alerted human only.

19. The method of claim 18, wherein determining whether the predefined condition is met based on real-time locations of the human and the autonomous vehicle comprises determining whether the predefined condition is met based on a predicted path of movement of the autonomous vehicle and/or the human.

20. A non-transitory computer readable recording medium storing a computer program product for preventing collisions between an autonomous vehicle and a human, the computer program product comprising program instructions which, when run on processing circuitry of a collision prevention system, causes the collision prevention system to: determine based on real-time locations of a human and an autonomous vehicle, whether a predefined condition for an anticipated collision of the autonomous vehicle with the human is met; lock the autonomous vehicle if the predefined condition is met; alert the human for whom the predefined condition for colliding with the autonomous vehicle is met; wherein a subsequent unlocking of the autonomous vehicle is performable or initializable by the alerted human only.

Description

BRIEF DESCRIPTION OF THE DRAWINGS

(1) These and other aspects of the present disclosure will now be further described, by way of example only, with reference to the accompanying figures, wherein like reference numerals refer to like parts, and in which:

(2) FIG. 1 shows a schematic illustration of a collision prevention system according to some example implementations as described herein;

(3) FIG. 2 shows an unlocking communication scheme according to some example implementations as described herein;

(4) FIG. 3 shows a flow-diagram of a method according to some example implementations as described herein;

(5) FIG. 4 shows a block-diagram of a collision prevention system according to some example implementations as described herein; and

(6) FIG. 5 shows a localization system and a collision determination unit according to some example implementations as described herein.

DETAILED DESCRIPTION

(7) The present disclosure may be applied to collaborative collision avoidance for, for example, industrial autonomous vehicles. The combination of the tag, anchor and platform with the locking/unlocking mechanism with advanced intelligence according to some example implementations as described herein may be particularly advantageous in order to avoid collision between autonomous vehicles and humans.

(8) It has been realized by the inventors that fixed trajectory solutions are not compatible with dynamically changing environments. Furthermore, collision avoidance may make use of a real-time localization system for following moving objects with category/identity features. However, their precision may not be good enough in high-risk decisions, and they may not be able to replace the scanning sensors.

(9) In view of the lack of a suitable detection method, the necessity of human driver assistance makes autonomous driving not entirely autonomous and it is not cost efficient either.

(10) The location intelligence solutions as described herein may overcome these problems. In particular, the location intelligence solution as described herein may allow for collision avoidance of, for example, workers and autonomously guided industrial vehicles in dynamically changing environments. Real-time position information of both workers and vehicles with identities may be used. This way, the vehicle navigation algorithm may be able to use the assistance of the workers impacted, instead of, for example, the assistance of an additional human driver.

(11) In some example implementations, the following steps may be taken in some scenarios. First of all, a near-collision situation may be detected probabilistically based on determined real-time locations of a vehicle and a human (e.g. worker). A conservative approach towards locking of the vehicle may then be taken. Impacted workers may thereafter be alerted. Finally, subsequent unlocking of the vehicle may be performed by those impacted workers only.

(12) Centralized and distributed variants are presented throughout the present disclosure. Using direct proximity measurements and communication between vehicle and workers in the latter may help overcoming limitations of global solutions.

(13) The collision prevention system as described herein may be used by autonomous vehicles and workers in collaborative work areas for assisted collision avoidance in dynamic environments. A secure communication infrastructure with known characteristics may need to be provided according to some example implementations.

(14) FIG. 1 shows a schematic illustration of a collision prevention system 100 according to some example implementations as described herein.

(15) In this example, the collision prevention system 100 comprises a location intelligence system 116. The location intelligence system 116 may be used for early collision detection and communication to vehicles and workers.

(16) In some example implementations, the location intelligence system 116 may be proprietary. Alternatively, the location intelligence system 116 may integrate or be integrated in an existing intelligence system of an industrial monitoring system.

(17) The location intelligence system 116 may, in some examples, provide feed for existing industrial monitoring or traffic control systems.

(18) In this example, the location intelligence system 116 makes use of (i) maps with dynamic zone definitions of working areas, (ii) real-time position streams from the localization system, which is, in this example, an indoor localization system 112, (iii) a historical database for optional machine learning, and (iv) additional sensorial data from an optional industrial monitoring system.

(19) In this example, the location intelligence system 116 is coupled to the indoor localization system 112.

(20) The indoor localization system 112 may be used for positioning of both workers and vehicles, or direct proximity measurements between workers and vehicles.

(21) In this example, the indoor localization system 112 comprises tags 110a and 110b, anchors 102a, 102b and 102c, and a platform which comprises, in this example, localization platform cloud services 114.

(22) In this example, the anchors 102a, 102b and 102c are in communication with the localization platform cloud services 114. Data obtained via the anchors 102a, 102b and 102c regarding the position of the vehicle 104 and the human 106 may be provided to the localization platform cloud services 114 where the data may be processed in the cloud.

(23) The anchors 102a, 102b and 102c may be building blocks of the localization infrastructure, performing the positioning measurements and communication with tags 110a and 110b and the platform. The anchors 102a, 102b and 102c may have fixed known positions for absolute positioning of the centralized solution and/or be mobile, e.g. installed on the vehicles for relative proximity measurement of worker tags 110b.

(24) As outlined above, the anchors 102a, 102b and 102c are further in communication, in this example, with the tags 110a and 110b. This may allow for improved position determination of the vehicle 104 and the human 106, which/who may be uniquely identified via the tags 110a and 110b.

(25) The tags 110a and 110b may be (small) devices fixed to workers and/or vehicles. They may be positioned and may be associated to a unique ID and may, in some examples, use a battery for power supply.

(26) The platform (in this example the localization platform cloud services 114) of the indoor localization system 112 may be the central system, providing localization services. It may be responsible for communicating with the anchors 102a, 102b and 102c, performing collection of raw measurements, post-processing, and providing a real-time location stream to the location intelligence system 116.

(27) In this example, the collision prevention system 100 further comprises connectors 108a and 108b which are in communication with the location intelligence system 116. The connectors 108a and 108b may be used for alerts and command actuation.

(28) In this example, locking/unlocking triggering commands may be sent from the location intelligence system 116 to the connector 108a. The connector 108a may hereby be used to deliver locking/unlocking commands to the navigation system which is used to control the autonomous vehicle, or to the autonomous vehicle directly. Furthermore, the connector 108b may be used by the human 106 in order to send an OK message to the location intelligence system 116, signaling that the human 106 is safe. Furthermore, the location intelligence system 116 may, in this example, send an alarm (alert) to the connector 108b. The connector 108b may, in some examples, comprise a mobile phone application.

(29) As shown in FIG. 1, the location intelligence system 116 may, in some examples, report and be integrated into an industrial monitoring system 118.

(30) In this example, the indoor localization system 112 has two parts: the platform and the physical devices, e.g., tags on pedestrian workers (110b) and vehicles (110a), anchors infrastructure (102a, 102b and 102c) installed on a mine tunnel wall and the connectors 108a and 108b which may be a personal device and an engine controller, respectively.

(31) The anchors 102a, 102b and 102c may send their raw measurements to the localization platform cloud services 114. Direct proximity measurement may be performed if tag 110a is replaced by a moving anchor. Location intelligence system 116 may handle a near-collision situation detection from localization input and navigation planning, sends alarms and locking triggers to the connectors 108a and 108b, and handles OK messages from the replying uniquely identified endangered workers. Finally it will only send an unlocking trigger to the vehicle 104 when the dangerous situation is dissolved completely.

(32) As outlined above, the system may also report and be integrated into an industrial monitoring system 118. Although the location intelligence system 116 may be necessary for this logic, it is indicated in FIG. 1 that it may be part of the existing industrial monitoring system 118.

(33) Methods as described herein may be adapted to global infrastructure-based or direct-measurement-based variants of the system and use various worker unlocking signals. The latter may, in some examples, be a fallback method used in parallel with the global infrastructure.

(34) FIG. 2 shows an unlocking communication scheme 200 according to some example implementations as described herein.

(35) Different example implementations can be deduced from FIG. 2.

(36) In some examples, a common part of the different example implementations of the methods may be that once a near-collision situation is detected (i.e. a collision may be predicted if no measures are taken to change, for example, a predetermined path of the autonomous vehicle), a lock-down may be initiated in the navigation of the autonomous vehicle. Furthermore, a local danger zone may be defined conservatively, considering one or more of (i) compensation for uncertainty in localization measurements, (ii) a reaction time of a closed loop vehicle control, (iii) the planned trajectory of the vehicle, and (iv) positions and velocities, and learned movement characteristics of the (near-by) workers. Movement characteristics may hereby, in some examples, refer to common speeds (for example average speeds) of movement of works and/or direction changes of workers.

(37) Common to the different example implementations may further be to send an alert signal to the identified worker(s) located within the danger zone. Furthermore, the vehicle may be unlocked once each involved work has taken, for example, at least one of the following actions: signaling their safety after relocating outside of the reach of the autonomous vehicle, and leaving the danger zone (which may be defined with the lock-down alert).

(38) The different example implementations may vary based on positioning and defining local danger zones. On the one hand, absolute positions as measured, for example, by a pre-installed infrastructure, planned trajectory information, historical datasets and available communication channels between anchors and the localization platform may form the basis for positioning and defining local danger zones. Alternatively, only the relative distance, e.g. proximity measurements as measured by a moving anchor on the vehicle and its velocity vector, using direct communication between the moving anchor and the worker tag (e.g., activating a connected alert device on the worker), and the moving anchor and the vehicle connector may be taken into consideration when positioning and defining local danger zones.

(39) The different example implementations may further vary based on the types of worker safety/unlocking signals. In some examples, pressing of a physical button and/or reading of, for example, the unique ID badge of the worker by a separate device may be used as such signals. Alternatively (or additionally), a mobile device application may send a safety signal to the location intelligence system that may send an unlocking signal to the vehicle when no more endangered workers are detected. Furthermore, alternatively (or additionally), detection of a worker leaving the danger zone may cancel an alert.

(40) In any case, a second level of danger zone may be defined by the location intelligence system. Unlocking may only be done when no workers are located within this ‘red’ zone.

(41) In the example depicted in FIG. 2, three endangered workers and one vehicle are considered in this scenario.

(42) All locations may be periodically measured and known in the central application via the localization platform. Once the central logic of the location intelligence system detects a high probability of collision (i.e. higher than a predetermined threshold) or generally that the predefined condition for an anticipated collision between the autonomous vehicle and the worker(s), the vehicle is locked, and all endangered workers receive alerts.

(43) The vehicle only becomes unlocked after all three workers signaled their own safety via one or more of (i) pressing a button on the stopped vehicle where they are identified by their unique location tag ID, (ii) sending a signal on a mobile phone application, and (iii) exiting the conservatively defined danger zone.

(44) A higher level of security may be achieved by utilizing a second level danger zone. Unlocking may require a worker signal from outside of the second level danger zone.

(45) FIG. 3 shows a flow-diagram of a method 300 according to some example implementations as described herein.

(46) In this example, it is determined at step S302, based on real-time locations of a human and an autonomous vehicle, whether a predefined condition for an anticipated collision of the autonomous vehicle with the human is met.

(47) The autonomous vehicle is then locked at step S304 if the predefined condition is met.

(48) The human for whom the predefined condition for colliding with the autonomous vehicle is met is then alerted at step S306.

(49) In some examples, unlocking of the autonomous vehicle may be performed or initialized at step S308 by the alerted human only.

(50) In order to realize the above-identified example implementations, a collision prevention system 100 as shown in FIG. 4 is provided.

(51) In this example, the collision prevention system 100 comprises a localization system 402 configured to determine positions of an autonomous vehicle and a human.

(52) The collision prevention system 100 further comprises a collision determination unit 404 coupled to or in communication with the localization system 402, whereby the collision determination unit 404 is configured to determine a predefined condition for an anticipated collision of the autonomous vehicle with the human is met based on the determined positions (and/or the predicted trajectories) of the autonomous vehicle and the human.

(53) In some examples, the localization system 402 comprises the indoor localization system 112. Additionally or alternatively, the collision determination unit 404 may comprise, according to some variants, the location intelligence system 116.

(54) As shown in FIG. 5, the localization system 402 may be partially cloud-based, as indicated by the cloud 502. In this example, the localization system 402 comprises a processor 504 and memory 506. Furthermore, in this example, the processor 504 and memory 506 are comprised in cloud 502.

(55) In this example, the collision determination unit 404 is cloud-based, as indicated by cloud 508. In this example, the collision determination unit 404 comprises a processor 510 and memory 512 which are, in this example, comprised in cloud 508.

(56) As will be appreciated, cloud 502 and cloud 508 may, in some examples, be integral to a single cloud. Additionally or alternatively, processor 504 and processor 510 may be integral to a single processor. Additionally or alternatively, memory 506 and memory 512 may be integral to a single memory.

(57) Multiple locations may deploy separate physical infrastructures, and may still use central cloud micro-services. Industrial safety use-cases, however, may have their own dedicated cloud infrastructure for low delay, near real-time communication. The choice may depend on the speed of vehicles and the remoteness of the location in which the system may be implemented.

(58) As outlined above, the location intelligence as described herein may be used in indoor industrial settings. By using indoor localization infrastructure for both workers and autonomous vehicles, their collaborative interaction in case of danger of collision may be used for unlocking of a vehicle slowing down or stopping once the danger has passed. Perfect autonomous navigation is not replaced by assistance of a human driver, but by assistance of the endangered pedestrians. This may be achieved by centralized and direct communication variants.

(59) Variants and example implementations of systems and methods as described herein may enable use of autonomously guided vehicles without human driver assistance in high-risk zones, e.g., collaborative and dynamic work areas in a factory or mine.

(60) Systems and methods as described herein may further be used in indoor industrial areas.

(61) The indoor localization as described herein may replace methods like environment scanning (LIDAR or image processing).

(62) The solution described herein enables use of probabilistic location measurements for collision avoidance, as worker assistance may compensate missing information.

(63) Situation awareness and reactive navigation decisions may avoid under-optimal general speed restrictions. For example collaborative work areas with a low number of workers may not have the same slow-down requirements.

(64) Example implementations as described herein may further allow easy installation of infrastructure and provide the option of direct proximity measurements.

(65) Examples as described herein may be provided as a complete solution, or may be integrated into an existing location intelligence and monitoring system.

(66) No doubt many other effective alternatives will occur to the skilled person. It will be understood that the present disclosure is not limited to the described variants and encompasses modifications apparent to those skilled in the art and lying within the scope of the claims appended hereto.