[Sep 18 – Apologies for the belated posting. Here’s the update from Daniel. – Eric]
We continued discussions on the mobile navigation base drive system and reviewed past PyroTech robots for inspiration. The next step is to draw up some design ideas and work out the interfaces between the drive system and the superstructure to ensure we have the right dimensions for everything to fit in place.
We also expanded our investigation into the RatSLAM video-based mapping software. Previously we had demonstrated that the software worked using video provided by the original author. This time we used an Android smartphone program to capture images from a phone and send them to the software. This allowed us to see how the software works when mapping The Forge HQ shop. It was clear that the software is able to get the topology of the environment, but that the poor quality of the video-based odometry made the map unreliable and unrealistic. We discussed plans to improve the odometry by using wheel encoders and an inertial measurement unit (IMU) in the drive base. Next step for RatSLAM is to see if we can tune some of its parameters to understand it better and improve the performance.
It has been a fairly busy week. On Friday the 1st, the Structure team met to decide on an approach for building the Dalek base. We settled on building a shell out of sheet metal, stabilized and reinforced by internal shelving and mounting hardware for the other components. The “foot” of the Dalek will serve as a mounting point for the drive module as well as an array of distance/obstacle sensors. For internal access (to batteries, computers and such), the “back” of the shell will be removable. We presented this plan to the rest of the group at tonight’s meeting, and got the thumbs-up to move forward. After taking measurements of the KEN robots, we got some scaling figures, and will be working on a mock-up for sizing the rest of the components in the coming weeks.
The Drive team continued their discussions tonight, and after reviewing several options, decided to focus on a 4x omniwheel-based holonomic system. We’ll be shopping around to get some idea of pricing, and then moving on to power and control questions.
Wrapping up tonight’s meeting, the Software team made some exciting progress. After installing OpenRatSLAM on KEN4’s main computer, they were able to run it and observe the algorithm in action, albeit using prepackaged data.
That’s all for today. We made good progress, so I think we can sleep soundly knowing that our Dalek overlord will be pleased.
At tonight’s meeting, the Drive and Sensor teams began to move from the requirements we worked out last week, to early design sketches.
The Drive team began with some exploration of existing drive systems, with a priority of keeping cost and complexity low. Our preliminary requirements call for a simple differential drive system, which provides linear motion and turning. However, we were reluctant to give up lateral movement capability. If possible (and practical), we’d like to adapt a proven drive system to connect with a standard interface of our own design. This modularity would permit us to swap drive systems easily, enabling future upgrades as well as potentially helping us to meet our single-part weight constraints. If we do not find a suitable existing drive system to adapt, we’ll consider building either a two-swerve + two-caster system, or a ball drive.
On the Sensor team, specific types of sensors were selected, and software packages identified that can leverage them. The current plan includes ultrasonic and IR sensors for distance measurement and proximity detection, Raspberry Pi cameras, low-cost LIDAR, and a depth camera. These would feed into our own control software, as well a stereo vision module, OpenRatSLAM for visual odometry and the gmapping package for ROS.
The Structure team will meet later this week to discuss physical design. Yet to be discussed are Computer requirements, but those will become clearer after the Software and Sensor teams refine their designs.
Hello again! Yes, there has been another lapse between updates (and this author is partly to blame), but work has still been proceeding.
Recent meetings have been dedicated to repairs and maintenance on the robots (plus adding a slow blink), and earlier this month, Daniel gave a presentation on the project to the local IEEE Robotics and Automation group. It was here that the idea to place KEN onto a mobile base resurfaced.
At tonight’s meeting, we ended up in discussions about motivating ideas for the project, and the mobile base turned out to be quite popular. While interesting (and still on the to-do list), it seems that game-playing just isn’t quite as compelling as autonomous movement and navigation on a robot’s feature list.
Having settled on this for our next major goal, we spent the rest of the meeting working out requirements. The first one, of course, is that it has to look like a Dalek. (But don’t worry, “does not actually exterminate” is also a requirement – it was suggested that we call it “The Friendly Dalek.”)
The addition of mobility will open up many new possibilities for the humanoid robot project. It will also take a lot of (fun!) work. There is plenty of room (and need) for more people to join us, so we hope to see you at our next meeting!
Robots in repose: a semi-disarmed KEN v2 waits patiently for its other limb, while KEN3 meditates on cable management.
This week, we continued our work from the last meeting, and all but one of our robots got some attention. Repairs were made to KEN 1’s neck, and the robot readied to go out for another demo. KEN v2 received the same NTP (time server) setup that KEN 3 & 4 got last week, after removing and temporarily replacing an unreliable USB power brick.
KEN4’s head is coming together, and a 3D-printed pulley was added to the neck assembly, providing an attachment for a spring or other tensioner that will ease the burden on the tilt servo. The same modification will be added to KEN3, which was undergoing vision system tests this time.
Next time, we’ll continue testing KEN3, and start putting the last pieces in place for KEN4’s structure and mechanisms.
PyroKEN getting new songs on one side of the tables, and synchronized clocks on the other. Can you guess which is which?
We managed to check a few more items off of our to-do list this week. Network Time Protocol (NTP) servers were set up on the main computers of both PyroKEN and KEN4 (KEN1 replicas), which resolves an issue that crops up in the vision system when no Internet access is available. The broken tilt servo in PyroKEN’s neck was replaced as well. With a little bit of counterbalance and calibration, PyroKEN’s mechanisms should be functionally complete. Progress on KEN4’s neck and hardware mounting also moved forward, and it isn’t far behind its sibling.
On another note, PyroKEN was “taught” a few new songs. (The default AI script contains only one.) Now, we have even more ways to enjoy KEN’s soothing, monotone voice for awkwardly long spans of time. (A “stop” command was hacked in, to the great relief of those present.)
It was a productive meeting. We’re getting really close to having two more operational KENs, and excited to start working on enhancements.
Kevin prepares PyroKEN’s head. (We may want to make sure the robots don’t see this.)
It has been a while since the last update, but work has still been progressing. The project is now working on several robots: refined replicas of KEN v1, a dissected KEN for educational display (“Flat KEN”), and of course KEN v2.
Most of the work recently has been directed toward finishing the replicas and Flat KEN. This week, the team got nearly all the systems working in PyroKEN (with the exception of a neck servo, which only needs to be swapped out). At our next meeting, we’ll continue on with the mechanical work, and add enhancements to the software systems to make them more robust and user-friendly.
When the robots are all complete, we’ll have several platforms with which we can begin exploring the next phase of the project: making the robots play games. More on this soon, along with additional information on our other robots.