Industrial Robot Fleet Management System

Transcript

1. Jean-Marc Alkazzi – Applied AI Lead Industrial Robot Fleet Management System
2. Industrial Robot Fleet Management System
3. Industrial Robot Fleet Management System
4. Industrial Environments: Expectations Industrial Environments 4 https://www.belvoir.net/wp-content/uploads/2022/10/iStock- 1138429558-scaled.jpeg https://www.wealthmanagement.com/sites/wealthmanagement.com/files/styles/ article_featured_retina/public/co-warehousing.jpg?itok=xEH8z4Yk https://alta-commercial.com/wp- content/uploads/2020/06/Warehouse-Building.png
5. Industrial Environments: Reality Industrial Environments 5 Inside a BMW Group Production Environment
6. Industrial Robot Fleet Management System
7. iw.hub 7 Autonomous Navigation 2.2 m/sec top speed Dynamic Obstacle Avoidance
8. iw.hub 8 This is what we call an industrial Chassis
9. iw.hub 9
10. Safe with Magnetic Sensors Lift objects from 222mm height Lift up to 1000 Kg! iw.hub 10
11. iw.hub 11 Extensive Ports Powered by NVIDIA® Jetson AGX Xavier Industrial Casing
12. >8hrs on 1 charge BMW i3 Battery Refurbished Battery iw.hub 12
13. 2x LiDARs RGB-D Camera Human-Robot Interaction iw.hub 13
14. Redundancy and Diversity of Safety Safety Switches CE Certified iw.hub 14
15. Industrial Robot Fleet Management System
16. Industrial Robot Fleet Management System A Group of Robots
17. Robot Fleet: Challenges Robot Fleet 17 1. Which robot should do which task? 2. Which route should each robot take? 3. Can the robots autonomously overtake each other? 4. When should we trust the robot's local decisions? 5. What to do in case of sudden congestion? 6. What if the robots can't resolve a deadlock on their own? 7. How can we optimize the overall travel distance during task assignment? 8. If a robot goes into a long narrow road, how can another robot know to not go in from the opposite direction if they can't even see each other? 9. How can robots from different vendors communicate to collaborate? 10.How should a robot know that a fire alarm is on, and it should go to a pre-defined safety location without causing congestions with others? 11.….
18. Robot Fleet: Challenges Robot Fleet 18 1. Which robot should do which task? 2. Which route should each robot take? 3. Can the robots autonomously overtake each other? 4. When should we trust the robot's local decisions? 5. What to do in case of sudden congestion? 6. What if the robots can't resolve a deadlock on their own? 7. How can we optimize the overall travel distance during task assignment? 8. If a robot goes into a long narrow road, how can another robot know to not go in from the opposite direction if they can't even see each other? 9. How can robots from different vendors communicate to collaborate? 10.How should a robot know that a fire alarm is on, and it should go to a pre-defined safety location without causing congestions with others? 11.….
19. Industrial Robot Fleet Management System
20. Industrial Robot Fleet Management System A Group of Robots
21. Industrial Robot Fleet Management System AnyFleet
22. Industrial Robot Fleet Management System AnyFleet I do way more than just manage robots!
23. Industrial Robot Fleet Management System [Simply] Fleet Manager [Simply] 23 Task Assignment • For each triggered Task, assign a robot to accomplish it. • Track Tasks to figure out when they are completed. • Respect different rules per factory. Route Planning • For each robot, propose the best possible route to avoid congestions. • Track robot movements and factory state to update routes accordingly. • Support robots which can generate their own route (AMRs) as well as those which can't (AGVs). Charging Strategy • Sustain a high throughput of accomplishing missions by charging robots just when necessary. • Execute a charging strategy which extends the robot's battery lifetime. • Charge robots in preparation for expected high throughput.
24. How Hard can it be?
25. Build the most advanced Fleet Management System to coordinate the most complex logistics processes at scale. The Vision
26. Why is a Fleet Manager even needed? Can't we just have smarter robots and solve the problem? The Existential Question
27. Aren't Smarter Robots enough? Fleet Manager [Simply] 27 In Some cases, maybe yes. In Some other cases, not really. How do you adapt based on situation, robots, and external factors? Especially with our iw.hub! Even with the above "human- level" intelligence.
28. idealworks' Approach idealworks' Approach 28
29. idealworks' Approach idealworks' Approach 29
30. idealworks' Approach idealworks' Approach 30
31. idealworks' Approach idealworks' Approach 31
32. idealworks' Approach idealworks' Approach 32
33. idealworks' Approach idealworks' Approach 33
34. idealworks' Approach idealworks' Approach 34
35. Industrial Robot Fleet Management System AnyFleet
36. Industrial Robot Fleet Management System [Simply] Fleet Manager [Simply] 36 Task Assignment • For each triggered Task, assign a robot to accomplish it. • Track Tasks to figure out when they are completed. • Respect different rules per factory. Route Planning • For each robot, propose the best possible route to avoid congestions. • Track robot movements and factory state to update routes accordingly. • Support robots which can generate their own route (AMRs) as well as those which can't (AGVs). Charging Strategy • Sustain a high throughput of accomplishing missions by charging robots just when necessary. • Execute a charging strategy which extends the robot's battery lifetime. • Charge robots in preparation for expected high throughput.
37. Industrial Robot Fleet Management System [Simply] Fleet Manager [Simply] 37 Task Assignment • For each triggered Task, assign a robot to accomplish it. • Track Tasks to figure out when they are completed. • Respect different rules per factory. Route Planning • For each robot, propose the best possible route to avoid congestions. • Track robot movements and factory state to update routes accordingly. • Support robots which can generate their own route (AMRs) as well as those which can't (AGVs). Charging Strategy • Sustain a high throughput of accomplishing missions by charging robots just when necessary. • Execute a charging strategy which extends the robot's battery lifetime. • Charge robots in preparation for expected high throughput.
38. Task Assignment: The Goal Task Assignment 38 Assign the right task to the right robot, to accomplish them at the correct time, in the correct sequence while minimizing total travel time.
39. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 39 20min / / 60min Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday.
40. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 40 20min / / 60min 25min / / 45min Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday.
41. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 41 20min / / 60min 25min / 28min / / 45min / 28min Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday.
42. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 42 20min / / 60min 25min / 28min / / 45min / 28min Caused by the limited number of tools. Consequence of trying to parallelize. Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday.
43. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 43 20min / / 60min 25min / 28min / / 45min / 28min Both can be improved, for instance by cooking in different order. (e.g., prepare the fries while someone prepares the salad, so you don't wait for the only salad bowl to be available) Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday. Caused by the limited number of tools. Consequence of trying to parallelize.
44. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 44 20min / / 60min 25min / 28min / / 45min / 28min Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday.
45. Task Assignment: Simplified Analogy (Cooking Schnitzels) Task Assignment 45 Imagine yourself in an apartment with a small kitchen trying to cook Schnitzels on a Sunday. 20min / / 60min 25min / 28min / / 45min / 28min There is a limit on how much you can parallelize in a given kitchen.
46. Task Assignment: Challenges Task Assignment 46 In the Kitchen 1. Know what can and what should be parallelized given the current circumstances. 2. Take into consideration all the "shared tools" to parallelize efficiently. 3. Understand the limits of your current environment. In the Warehouse [Same as above +] 1. Solving the problem should be very fast, otherwise the factory state would have changed by the time you want to execute your solution. 2. Respect rules specific to given factories (e.g., no 2 close-by dollies should be assigned to robots simultaneously). 3. Possible combinations of assigning tasks to robots becomes exponentially bigger with more robots/tasks.
47. How Hard can it be?
48. Task Assignment: Challenges Task Assignment 48
49. Task Assignment: Challenges Task Assignment 49
50. Task Assignment: Challenges Task Assignment 50
51. Task Assignment: Challenges Task Assignment 51 Number of ways to assign all 3 robots to 3 tasks is 3P3 = 6
52. Task Assignment: Challenges Task Assignment 52 We didn't account for robots not assigned to anything...
53. Task Assignment: Challenges Task Assignment 53 If we do, they end up being 34 possibilities.
54. Task Assignment: Challenges (Pop Quiz) Task Assignment 54 . . . . . . 10 Robots 30 Tasks There would be nearly - 500 Possibilities - 30 000 Possibilities - 3 000 000 Possibilities
55. Task Assignment: Challenges (Pop Quiz) Task Assignment 55 . . . . . . 10 Robots 30 Tasks There would be nearly - 500 Possibilities - 30 000 Possibilities - 3 000 000 Possibilities 172 890 164 557 291 Possible Assignment Combinations ~173 Trillion Combinations!!
56. Task Assignment: Challenges (Pop Quiz) Task Assignment 56 . . . . . . 10 Robots 30 Tasks There would be nearly - 500 Possibilities - 30 000 Possibilities - 3 000 000 Possibilities 172 890 164 557 291 Possible Assignment Combinations ~173 Trillion Combinations!! Fun Fact: If you spend $1 Million per day, you need ~470 000 years to spend $173 Trillion
57. Task Assignment: AnyFleet Task Assignment 57 1. We solve all the previously mentioned challenges, and even bigger scales. 2. We respect different custom rules per customer. 3. Optimize Overall Distance Travelled by all robots. 4. We support multi-fleet multi-robot-types during the same assignment request. 5. Schedule tasks in advance for a specific robot about to finish a mission (The Uber-like feature). 6. and much more!
58. Industrial Robot Fleet Management System [Simply] Fleet Manager [Simply] 58 Task Assignment • For each triggered Task, assign a robot to accomplish it. • Track Tasks to figure out when they are completed. • Respect different rules per factory. Route Planning • For each robot, propose the best possible route to avoid congestions. • Track robot movements and factory state to update routes accordingly. • Support robots which can generate their own route (AMRs) as well as those which can't (AGVs). Charging Strategy • Sustain a high throughput of accomplishing missions by charging robots just when necessary. • Execute a charging strategy which extends the robot's battery lifetime. • Charge robots in preparation for expected high throughput.
59. Route Planning: The Goal Route Planning 59 Assign a route per robot so that each robot reaches its destination in the fastest time possible while avoiding potential congestions
60. Route Planning: Simplified Analogy (Where to go?) Route Planning 60
61. Route Planning: Simplified Analogy (Where to go?) Route Planning 61
62. Route Planning: Simplified Analogy (Where to go?) Route Planning 62 ?
63. Route Planning: Simplified Analogy (Where to go?) Route Planning 63
64. Route Planning: Simplified Analogy (Where to go?) Route Planning 64 Map of Munich, with traffic on routes, with idealworks pin.
65. Route Planning: Challenges Route Planning 65 1. Time Delays caused by internet connection, not stable everywhere throughout the factory. 2. Execution Delays caused by different types of robots having different reaction times. 3. Unexpected Slowdown caused by traffic. 4. Roadblocks which requires a new path to be planned. 5. Chaotic Peak hours where all your predictions won't work. 6. Heterogeneous fleets (Older AGV fleet with new AMR fleet in same location).
66. Route Planning: AnyFleet Route Planning 66 AMRs or AGVs?
67. Route Planning: AnyFleet Route Planning 67 AMRs or AGVs? and Provide Route with different resolution levels. (e.g., go from A to B, or go from A to B passing through X, Y, Z) Use the Generated route from the edge device. Provide Route with higher resolution level.
68. Route Planning: AnyFleet (AGVs early preview) Route Planning 68 AGV provided routes on a corridor/intersection situations.
69. Route Planning: AnyFleet (AGVs early preview) Route Planning 69
70. AnyFleet Extensions: AnyTime (A Sneak Peek ) Route Planning 70 Showing “AnyTime”, part of our AnyFleet debugging suite extension, helping to replay any situation in any factory at any time. Replay our R&D Warehouse, forcing possible deadlocks by increasing density of robot in small areas, pushing our solution to its limits to find and solve edge cases.
71. Industrial Robot Fleet Management System [Simply] Fleet Manager [Simply] 71 Task Assignment • For each triggered Task, assign a robot to accomplish it. • Track Tasks to figure out when they are completed. • Respect different rules per factory. Route Planning • For each robot, propose the best possible route to avoid congestions. • Track robot movements and factory state to update routes accordingly. • Support robots which can generate their own route (AMRs) as well as those which can't (AGVs). Charging Strategy • Sustain a high throughput of accomplishing missions by charging robots just when necessary. • Execute a charging strategy which extends the robot's battery lifetime. • Charge robots in preparation for expected high throughput.
72. Charging Strategy: The Goal Charging Strategy 72 Ensure the maximum availability of robots while warranting an extended battery lifetime.
73. Charging Strategy: Simplified Analogy (24/7 Call-Center) Charging Strategy 73 You have a team of 5 people in a call-center. You want them to be available 24/7. 1. Everyone starts at 6AM, stays awake until 11PM. What happens with the 11PM till 6AM shift? 2. You have 2 shifts, 3 people for the morning part, 2 for the night shift. Night shift people are always stuck in the night shift. 3. Alternate the sleeping schedules over the week to give each person the chance to be in either shifts. Works well for robots [adds redundancy], may mess up humans' sleep schedule.
74. Charging Strategy: Simplified Analogy (24/7 Call-Center) Charging Strategy 74 You have a team of 5 people in a call-center. You want them to be available 24/7. 1. Everyone starts at 6AM, stays awake until 11PM. What happens with the 11PM till 6AM shift? 2. You have 2 shifts, 3 people for the morning part, 2 for the night shift. Night shift people are always stuck in the night shift. 3. Alternate the sleeping schedules over the week to give each person the chance to be in either shifts. Works well for robots [adds redundancy], may mess up humans' sleep schedule.
75. Charging Strategy: Simplified Analogy (24/7 Call-Center) Charging Strategy 75 You have a team of 5 people in a call-center. You want them to be available 24/7. 1. Everyone starts at 6AM, stays awake until 11PM. What happens with the 11PM till 6AM shift? 2. You have 2 shifts, 3 people for the morning part, 2 for the night shift. Night shift people are always stuck in the night shift. 3. Alternate the sleeping schedules over the week to give each person the chance to be in either shifts. Works well for robots [adds redundancy], may mess up humans' sleep schedule.
76. Charging Strategy: Challenges Charging Strategy 76 1. How to balance between availability and charge level? 2. What should we do if a sudden peak of tasks happens? 3. How can we ensure that we don't damage the battery with bad charging patterns? 4. How can we place chargers in the factory for a more efficient workflow?
77. Charging Strategy: AnyFleet Charging Strategy 77 1. Charging Strategy per Robot Type. 2. Integrates with Task Assignment Strategy. 3. Ensure maximum availability given number of robots. 4. Efficient charger placement as part of Use Case Setup. 5. and more!
78. Industrial Robot Fleet Management System AnyFleet I do way more than manage robots!
79. How can I benefit as a Customer? 79