Asked
Resolved by DJ Sures!
I am slowly getting all the quirks with the T265 worked out by docking etc. We really need some distance sensing as well so we can get better detail of our surroundings. @DJ are there any plans in adding a distance sensor like a D435 (or D435i since we already have an IMU in the T265).
Just trying to work out what I should purchase in anticipation of some future depth sensing capabilities.
I am slowly getting all the quirks with the T265 worked out by docking etc. We really need some distance sensing as well so we can get better detail of our surroundings. @DJ are there any plans in adding a distance sensor like a D435 (or D435i since we already have an IMU in the T265).
Just trying to work out what I should purchase in anticipation of some future depth sensing capabilities.
I'm struggling with the D435 because of the poor reviews and discontinued status. The T235 was a pretty big risk to add support for because of the "Intel way of abandoning customers".
So far the best bet is to agree on a lidar that you would all like to see supported. The Kinect was a disappointment - and I had high hopes due to all of the blogs with people demonstrating great successes... until I realized we're dealing with a Boston dynamics type scenario; where there's one usable result out of a thousand takes.
Really, I'm not sure what direction we should be heading. If we invest time into supporting a sensor that doesn't work, you'll be knocking on my door asking "why not" and my only response is "hey, I didn't make it" LOL. Any feedback on what direction you'd like to go is encouraged to help us - I am learning toward lidar.
But remember, any distance scanning device pushes data into the NMS so it's not really a lot of work. You can use the Javascript NMS methods and push data into the NMS from any device yourself as well.
We have heard rumours that Boston Dynamics is using the Realsense D4XX on SPOT. We have also seen it in Unitree, Dogotix and ghost robotics. If these robot dogs do take off, I think intel may have a steady customer base moving forward and a reason to keep the product in their supply chain. LIDAR is great for a 360 view but if you are trying to get a robot to climb up or down a flight of stairs I assume you need a direct line of sight to area you are heading in.
What use cases do you see for the Boston dynamics robots?
In a conversation I had yesterday in discord, we now believe that every BD SPOT has 5 * Intel D430 modules. The D430 is the same module in the D435 but has a custom RGB sensor added. Today SPOT costs 75K but Hyundai purchased them for mass production. I would estimate actual manufacturing and parts costs would be < 5K and they will probably retail once commoditized for under 10K.
I believe BD use cases fall into 3 categories The Good, The Bad and The Ugly.
The Good. Unlike a wheeled robot or a drone, a legged robot can go into areas that are accessible only by walking. Although our world has been transformed to support accessibility standards as soon as the power goes out or an alarm goes off, the lifts stop working. If there is debris in the way or hazards to navigate our wheeled robot friends will no longer function. If you have ever flown a drown in doors this requires a significant amount of skill to pilot a large drone and inevidably results in a collision in confined spaces. Small nimble drones do not have the battery to last longer than ~ 20 minutes and can't interact with surroundings (open doors, windows or move objects etc) and this is why humans still perform tasks in these environments today.
The ability for a Legged robot to function in these spaces makes them ideal candidates for any places that are not safe for humans. First responders (Firemen looking for survivors in a burning building, Police responding to the scene of a shooting, Bomb squad inspecting a package) are good first examples where SPOT is currently being deployed. Hazardous environments such as chemical leaks, nuclear power plants, machine plants, oil rigs, mining etc all provide use cases for SPOT to perform a task where a human should not.
The Bad: At under 10K SPOT like any robot will replace jobs. The security guard who patrols a building, the field tech who monitors and inspects equipment in the field, the auditor who verifies assets etc. There is a whole range of tasks humans perform just by walking around and observing and interacting with their surroundings, recording data and making changes as required. Initially these roles will be performed in a semi autonomous fashion with human operators remotely monitoring multiple robots simultaneously (Think Exosphere) to eventually fully autonomous robot that through repetition and reinforced machine learning will be able to perform these tasks without assistance.
One could argue as these human operated jobs go away, people will find other high value tasks to perform that are more suitable to a humans skills set. I recall the initial concerns that computers were going to take our jobs away but we all now have jobs designing, programming, operating and repairing computers.
The Ugly: I don't want to go here but I think we need to be realistic and understand that people will always use technology for nefarious purposes. BD original R&D was funded by DARPA and I don't think their interests were exactly humanitarian in nature. Dogs of war are coming and those trucks full of soldiers that fight wars in cities today, will soon be replaced with trucks full of robot dogs that have guns strapped to their backs. The US economy is fuelled by war and that is not going to end soon. This sadly will be one of the largest markets for Robot Dogs. Our children who spend every waking moment playing video games where they remotely control avatars to kill their opponents, have now become experts in their field in remote controlling killing machines. They have also become completely desensitized when it comes to indiscriminately killing an opponent. Our future wars will be fought, won and lost by an army of highly skilled teenagers remote controlling war dogs who will also be inadvertently killing innocent people and children with little to no remorse. What ever happened to the 3 laws of robotics?
@DJ said
Did you look at the two I posted from Robotshop? (there is actually a 3rd slightly more expensive but more self contained one now, but it is the same SDK as one of the others YDLidar ----$159 instead of sub $100). I would say if you look at the SDK's of both and one would be easier to write the skill for than the other, we would all be happy to settle for it (the next least expensive is over $300. I don't think many of us are interested in that much for a single sensor). It seems that YDLidar is the more active vendor with several devices in multiple price ranges, and I believe they may all use the same SDK, so if it is useable, that might be the best option.Here are the links again:
https://www.robotshop.com/en/rplidar-a1m8-360-degree-laser-scanner-development-kit.html https://www.robotshop.com/en/ydlidar-x4-360-laser-scanner.html and the new one: https://www.robotshop.com/en/ydlidar-g2-360-laser-scanner.html
Alan
If i can help in any way i will be happy to. Well, thats my 2 cents...Have fun
Manual in support section: https://synthiam.com/Support/javascript-api/Navigation/updateScan