Archive for September, 2005

20:25 hrs – Why did CB stall and scrape the walls?

Friday, September 30th, 2005

Fontana, CA – The good news is we have plausible explanations for why CB stalled, and also why it was scraping the walls in the tunnel. The first one is due to some errors in sensors and the second one indicates a fault in our code. Work is in progress to rectify both.Here is a little more detailed explanation for the robotics enthusiasts.

Like every team, we use GPS for getting the location of the bot. The location is measured in terms of latitude, longitude, and height above sea level. The GPS data is fed into another sensor called Interial Navigation Sensor. We are using RT3102 from Oxford Technical Solutions. The RT, as we call it, is a pretty sophisticated equipment. It fills in the gap between GPS readings. While a GPS gives location information five times a second, the RTgenerates the same information, and more, 100 times a second. It fills the gap by using some heavy duty physics equipment called accelerometers, gyro, and compass. There is a little more detailed involved. The raw GPS data, as used in cars for navigation, is very imprecise, with an accuracy of about 3 meters (which is about 10 feet). With that kind of accuracy, there is no way we can drive in a 10 feet corridor. We input to the RT a ‘differential GPS’ signal using C-NAV from C&C Technologies.

As with any program, the quality of output depends on the quality of input, which stated as a computing principle reads ‘garbage in, garbage out’.

When the GPS data has error within some tolerance RT and C-NAV together to a wonderful job in cleaning up the error. But when the error is very high, due to some atmospheric conditions, or just simple sensor errors, they cannot keep up.

The reason that CB stalled was due to some spurious error in the data received by RT and CNAV. The net result of the spurious data was that within about 1/100th of a second the bot thought it had dropped by 17 meter. That is more than the height of some houses.

Errors like these are what makes this challenge a grand challenge. If all the sensors were perfectly precise, life would be very simple. The challenge is in coping with sensor errors.

This is the first time we have directly experienced this situation. Amit, our lead in processing sensor data, along with Pablo and Suresh have worked out a strategy for dealing with spurious spikes location data.

The reason why CB scraped the walls of the tunnel is a little more intense. The short explanation is that we did not very well take care of the condition when the bot has inadvertantly hit something. In other words, we sort of assume that our algorithm is perfect and the bot will not hit anything. But if it does hit something, as it did, under that condition the algorithm does not behave well. In other words, our algorithm is not fault-tolerant.

Pablo and Suresh, the team responsible for obstacle avoidance, along with Dr. Tony Maida have worked out a strategy to address the above situation.

The bus I rented yesterday is coming in very handy today. The bus is really a house, with a living room and bed room. The bed room has been converted into a ‘cave’. The three of them are locked in it. We open the door, slip in food and water, and shut it back. They will be working in the cave through 10pm, the time when the garages close and everyone is asked to vacate the property.

Tomorrow early morning we get a practice run. A practice run is not on the NQE track. DARPA has cut out about 200 feet x 100 feet rectangles in the parking lot. We get to bring our bots there and try out changes in our algorithm. Our practice run is scheduled around 8am, but could be sooner if some other team does not take there slot.

– Arun Lakhotia

11:23 hrs – Run 2: Cleared the difficult track, but stopped for no obvious reason

Friday, September 30th, 2005

Fontana, CA – This has been heart wrenching. In the second run, CB cleared the difficult part of the track, about 3/4th of the track, and just when we thought she was home free she stopped, for no apparent reason. There was no obstacle, or anything else. Just a quick turn ahead of it. We are analyzing the data to find out what happened.

Gallery from NQE Qualifying Run #2

We were up at 5:00am, in the Speedway at 6:00am, and lined up for launch around 6:45am. Ahead of us was Red Team’s H1lander and Cornell’s Spider. I did not track time, but H1lander was launced around 7:00am. It went through the course pretty effortlessly. it was surprise to see Spider in the chute. The team was having problems with the transmission yesterday and were scheduled to change the transmission. They probably decided to run with the busted transmission. Spider was launched shortly after H1lander completed. It cruised through well too, and was visibly faster than H1lander along the speed section of the track. But halfway into the speed section in stopped. It had to be disabled and pulled out, most likely due to transmission failure.

CB was next in line for launch. Since day before we have been working on fixing the reason why it got stuck on the bail of hay. We were anxious to see how the changes fared, more so because the changes had not been tested.

As CB’s emergency lights started flashing we, in the spectators area, knew the bot was ready to launch. When it siren turned on it indicated bot had been given the signal to launch. But the bot did not move. Could feel the anxiety level in the team picking up. We saw the DARPA launch operator turn around to talk to our launch team – Joshua, Suresh, and Pablo. That did not look like a good sign. Joshua was seen running down. Not a good sign at all. Then Pablo came on radio–we have our own radio network here, thanks to James Dugal–and said they had forgotten to turn on the engine. Now that is a pretty good reason. We all had a big sigh of relief. That problem was easy to solve.

With the engine on, it was sure easy for CB to move. It took off from the chute, made a left turn and approached a gate, representing ‘cattle guard’. A little bit of twist and turn, and it aligned itself with the entrance, and pressed on gas. From then on it cruised through the twists and turns, without touching even a single traffic cone. The track has around 50 pairs of traffic cone. Each pair, one on either side of the track, represent a gate in the fence. Running into a cone is counted against you, as is running around any cone.

After a few turns, is the speed section of the track. I am not good in measuring distances visually, but it appears to be two segments each around 100 meters long. The speed limit for this section is 45 miles/hour. CB was cruising at 17-20 miles/hour, which is pretty close to its top speed.

As it completed the speed section, and approached the section with bale of hay, I was holding my breath and watching. Along with other teammates, I was standing on an bridge over the track and had a view straight on the track, looking at the bot approaching. CB was perfectly aligned to the middle line. An indication the software changes were working well. This section has two parts, first part with bales of hay on both sides of the track about 20 feet apart, and the second part with the bales about 10 feet apart. This time around CB approached the narrower section very well. It did dither a bit inside the narrow section, but did not attempt to climb the bales.

Once CB cleared the hay section, we had enough time to take one quick breath. The next section was simulating a 100 feet tunnel. The intent is to check how the bot behaves when it loses GPS signals. While we have known about this situation, we have not had the opportunity to test it ourselves. So far as CB is concerned, this situation was a complete surprise for her.

CB entered the tunnel well, straight in the middle. A very good sign. But about 20 feet into the tunnel, CB turned left and headed into the cement barriers forming the tunnel. Video shot by Scott shows CB skidding against the cement barrier. As it came out of the tunnel she made a hard left turn and ate her first cone.

Eating the cone must have made CB feel better, for very soon after she regained composure and avoided a card parked in the middle of the track. Then she approached the switchback, some sharp S-curves, went through them fine. After the switchback is a weird obstacle course, representing rocks and bushes on both sides of the track, and covering some part of the track. CB went through this part very well.

At this point CB had completed the difficult part of the track, the part that had narrow areas and small obstacles. The remaining section has wider tracks and the obstacles are much bigger.

I was mentally getting ready to do a victory dance. Just then we saw CB stop. Seconds, very long seconds, went by and she was still stopped. Turned into minutes, she was stopped. That’s not good news. CB doesn’t think that hard. If she is stopped this long there must be something bad.

The launch team was brought in to retrieve CB.

There was some satisfaction still. We did make some progress having completed 3/4th of the course.

The team is now analyzing data to find out the reason for the failure. You can be assured we will be back, as always.

– Arun Lakhotia

No qualification run today, a busy day nonetheless

Thursday, September 29th, 2005

Fontana, CA – The track was moving a lot slower today, so we did not get called for a second ‘run’. In DARPA terminology, two ‘runs’ make one ’round.’ DARPA guarantees to offer one round per team, but is hoping each team would get three rounds. Final evaluation will be based on the cumulative results.Today we did not get any run. But we still had a pretty busy day.

Yesterday, the three McGyvers, Mural, Joshua, and Adrian, had noticed that the sprocket connecting for the transmission on the left side was a bit off, coming out. They had started working on repairing that, but could not complete the repair. They came in early today to work on it. (10/01/05: I had earlier called them three amigos. But I think its more the three McGyvers is more appropriate, given their ability to make things out of what most may consider junk. Hint: TV show called McGyver.)

Turns out that the small size and ambhibious nature of the vehicle makes it really hard to operate on. Its belly is closed, which means if you drop anything while working on it, it stays in the belly. Its small size the engine area is pretty well packed, which makes it hard to reach into small, cramped areas.

When used as an ATV, it has seats that can be removed to reach these cramped spaces. But we have stacked almost 1,000 pounds of metal and electronics, cut through the body to attach our gear, and have wires and fuel lines running through. It would take almost a day to remove all our gear, then put everything back, and the re-calibrate all the sensors and actuators.

So you can understand what happens when the team members dropped a socket when trying to fasten a bolt. It took them over three hours to get the socket out on top of the hour needed to fix the problem they had started to fix.

With our hands full dealing with mechanical issues, we had no reason to complain that we did not get another run. If we were not ready when offered a run, then DARPA is not obliged to give us another offer.

Incidentally, the heat today was really bad. The sun just burns through the skin. The dirt and sand in the air only makes the matter worse.

To make it a little more comfortable for the team, today we rented a 36 feet long bus. Its pretty plush inside with nice sofa, kitchen, fridge, shower, bathroom, and bed. Tomorrow we will fit the bus to make it our computer room.

Oh by the way, our second run is scheduled tomorrow early morning. We have to be ready to go around 6:45am.

– Arun Lakhotia

CajunBot has a good start, but gets stuck in hay

Wednesday, September 28th, 2005

Fontana, CA – CajunBot had her first NQE run today. The run started off very well. The bot went through cattle guard and gates (made from traffic cones). She cruised along straight stretches of roads, went up and down small hills, and took smooth curves.She also went through one section of track lined with bales of hay on both sides, and was doing well going through a second section which was narrower than the first. She was running very close to the boundary, and about five bales of hay before that track was to be completed, it ran into a bale. This put it out of balance, and as it tried to return to track, could not gain composure and went on the hay on the other side, this time climbing to the point its front and middle wheels were up in the air.

With three wheels in the air, CajunBot did not have enough traction to continue moving, and had to be disabled by the DARPA officials.

 Joshua guiding CB to Run 1
Photo by – Scott Wilson

 CB rolling in hay
Photo by – Scott Wilson

This was the first run. DARPA guarantees that each team will have ‘offer’ for at least two runs. With one run down, it leaves us with at least one more run.

Of the twenty teams that ran today, only six teams completed the track without any significant hitch. Other teams, like us, were either disabled due to running on obstacles or  did not attempt a run (losing a chance).

After the remaining 23 teams have completed their run, we get another round. At the pace DARPA is going, it looks like we may get a second run tomorrow.

Read more about the first day’s results at http://www.tomshardware.com/hardnews/20050929_125919.html

— Arun Lakhotia

CajunBot made it to the opening ceremony

Wednesday, September 28th, 2005

Ontario, CA – The opening ceremony started at 9:00am, and CajunBot made its entrance at 8:59am.We had the replacement chain in hand at 1am, the Speedway gates opened at 6am, Joshua, Adrian, Murali, and Adam were in the garage at 6:05am. From there it took the whole three hours to put the new chain in place.

The main reason it took that long is that Max IV is a real compact machine. There is barely enough space to reach the drive chain without removing the engine. In the normal machine removing the engine is not a big deal. But with the changes made to build CajunBot, removing the engine and putting all the pieces back together could easily be a days job.

Joshua’s skinny physique and very malleable hands came in handy, no pun intended. Adrian and Murali were working with him. They tried to twist and turn their hands in, but to no avail. I do not know why Adrian even tried. With a football line backer physique, you’d not expect him to have Cinderalla fingers. Murali is not longer the skinny graduate student he was a few years ago. You can sense prosperity all around him, and there is more to come with Espion Interceptor, the anti-spam appliance he has built gaining notoriety of beating similar products from the big dogs of the industry.

 
Joshua working on chain
at about 7:17am Pacfic Time
CajunBot amongst other bots
after the opening ceremony
 

All the bots were expected to ‘assemble’ at the grand stand by 8:30am. That’s about the time Joshua was able to put clip the chain. We conveyed to DARPA that the bot will be ready in 10-15 minutes. By 8:50am we cranked the engine and left the garage, with Joshua driving the bot.

At 8:59am CajunBot was standing next to the other bots. By looking at her you could not tell the ordeal she had gone through.

Knowing what it took us to get to this point, The Star-Sprangled Banner at 9:02am had a completely new meaning.

– Arun Lakhotia

Chain Reaction

Wednesday, September 28th, 2005

Ontario, CA — At about 8:30 PM Arun Lakhotia contacted the Chase Team with grave news about a mechanical failure on the CajunBot. The chain had failed and CajunBot was down. Without a spare chain, the team was disheartened and feared the end had arrived before the start.However, their was a glimmer of hope in the eyes of team members when the chase team leader Danny Majors arrived at TGIF restaurant to pick up the broken chain in route to Ramona, California to look at what we hoped was a compatible replacement. With failed chain in hand, we roared up the ramp of I-15 without any idea whether our race to Ramona, CA would put us back in the DARPA race. Even more, we hoped for the possibly we might even get the CajunBot to the opening ceremonies on the infield of the California Speedway Wednesday morning.

The Ramona connection wasn’t the only option in the plan. Even if we couldn’t get CajunBot into the opening ceremonies, we needed a plan that would enable us to qualify. Within minutes, we’d formulated four options that could get us back on track.

Danny reached for the phone and I cracked the hatch of my laptop and we started pecking away in search of a source for a replacement chain. Danny’s first call was to Tom Cain, owner of Cain’s Equipment Sales in Lafayette, Louisiana. It was a late call and Mr. Caine was quite hospitable. While Mr. Caine didn’t have a replacement chain in stock, he did provide the name and phone number of the ATV manufacturer.

This information pointed us to the manufacturer’s website and list of dealers. We immediately located two dealers in the Southern California area: 2 the Max ATVs in Ramona, CA and Tico Tech, Inc. in Altadena, CA. These two incredibly helpful dealers offered Team CajunBot two promising options.

We made first contact with Gina Hampton at 2 the Max ATVs. Initially, we didn’t believe she’d be able to help us. She didn’t have a chain in stock, but she did have a new Max IV. After some discussion, Danny and Gina worked out a deal in which she would give CajunBot the chain off of her new vehicle. This deal took several phone calls to develop.

Meanwhile, between calls with Gina, we’d also contacted Rafael Calvo at Tico Tech, Inc. in Altadena, California. Rafael was very accommodating and offered the chain from his personal ATV. He even offered to deliver his ATV to our mechanics in Ontario, where Team CajunBot could remove the chain.

Two other options were unfolding within the few short minutes following “the call”. Another Max IV ATV owned by MedExpress, a Team CajunBot sponsor, would also offer the prospect of a replacement chain. When Danny contacted the MedExpress team, the ATV was actually on its way out of New Orleans as part of the Hurricane Katrina relief effort. A MedExpress ambulance was towing it, and it was nearly home. The plan was to remove the chain on arrival, and then deliver it to a member of Team CajunBot who was flying out of Lafayette tomorrow (Wednesday) morning.

The final option was simple, to deliver the broken chain to a nearby motorcycle shop for repair.

Danny and I tag-teamed between web and phone for what amounted to about 30 minutes before a solution emerged from the four options.

While each option offered a solid chance for a replacement chain early on Wednesday morning, only one option delivered the chain early enough to get Team CajunBot on the infield of the Grand Challenge Opening Ceremonies.

A quick return call to Gina at 2 the Max ATVs put the wheels in motion…the wheels on the Chase Team truck. With laptop on board, Verizon wireless Internet card inserted, we mapped the route to Ramona using Google Maps. It would be a 100 mile trek to see a chain we only hoped would match. We stopped at TGIF, picked up the broken chain and hit I-15 with the pedal to the medal. As we raced to Ramona, Gina’s husband George was busy working to remove the chain on their new MAX IV.

In an effort to quicken our trip, we asked Gina if she could meet us at the El Norte Parkway exit. It was no surprise she accommodated our request, saving us about 1 hour drive time. This also meant that it cost her an hour on the road.

Gina, husband George and their son George (and their toy Poodle) arrived shortly after us. After introductions, we laid out the new chain beside the broken chain to compare the two. It was a match, and there was a great sigh of relief that rippled from Encondido to Ontario to Lafayette.

We delivered the chain to a renewed Team CajunBot, despite the fact it was 1:30 AM. The team plans to arrive at the CajunBot garage bay at 6:00 AM this morning to begin installation of the new chain. With a bit of luck and some hard work, you should see Team CajunBot beside CajunBot at opening ceremonies.

On behalf of Team CajunBot, we appreciate the sacrifice and good will offered by everyone who assisted us in solving this problem, especially the great people from 2 the Max ATVs.

Please join us in showing your support by visiting the 2 the Max ATVs website at www.2themaxatvs.com

— Mike Spears

Drive chain broken, Chase Team rushing to get one in middle of night

Tuesday, September 27th, 2005

Ontario, CA – We are in a SOS mode. Around 7pm, just when we were ready to retire early today, and were packing CajunBot in the trailer, its rear drive chain broke. The bot is currently not operational. By the time we left the Speedway, around 9:30pm, the status was that bot would not participate in the opening ceremony. By 11:00pm the Chase Team was on its way to San Diego to pick up a new chain, and there is hope that CajunBot will be ready for the opening ceremony at 8:30am.


Joshua, Adrian, and Murali inspecting the broken chain

Here is a peep at some of the happenings…

When I learned of the problem, I called Danny, as I always do in times of crisis. Danny called Brother Ray Majors, his dad, as he always does. And Brother Majors had a solution, as he always has. He has another MAX IV ATV (the first one has been converted into CajunBot). Bro. Ray was on his way to New Orleans, but turned around and went back to Melville, ready to pull out chain from his vehicle. If needed he will bring the chain to Lafayette airport early in the morning tomorrow to give to Patrick Landry, who flies out at 5:40am. While that is a good option, and we would have the chain by noon, we still wanted something sooner.

After a lot of frantic calling, around 10:30pm things started to change direction. Dr. Guna Seetharaman and Mike Spears, independently, located two MAX IV dealers within two hours driving distance.The surprising part is the dealers answered their phone at this late hour. They both were willing to take chain off their vehicles and sell it to us.

By 11:00pm Danny and Mike were on the road to San Diego. They are on their way as I write.

If all goes well, early morning at 6:00am, when the Speedway opens, Joshua, Murali, Adrian, and Adam will be at the garage to install the chain. By 8:30am we expect CajunBot to be in the opening ceremony.

If you do not see CajunBot in the opening ceremony, you know why she is not there. If you see CajunBot in the opening ceremony, you know how she got there.

– Arun Lakhotia

Truth Tabernacle Church – Testing area for CajunBot

Sunday, September 25th, 2005

Ontario, CA – The Majors continue to out do themselves. Several weeks ago I had shared with Danny that I was worried where we will test our bots in Fontana, CA. He told me, “Don’t worry Doc, I’ll look into it.” Danny talked to his dad, who talked to his friend, who remembered so and so, son of a pastor in Brazil, married to a lady in Melville, whose so and so is a pastor in a church which is, guess what, four miles from California Speedway.Thanks to Brother Frazier, Team CajunBot has setup a camp in Truth Tabernacle, Bloomington, CA. The pastor and the church has been extremely gracious to the team. The parking lot of the church and an additional three acres property adjacent to the lot is available for us to test.

– Arun Lakhotia

Bots and Team regroup in Ontario

Saturday, September 24th, 2005

Ontario, CA – The bots and the team have regrouped in Ontario, CA, pretty much as per schedule. Had to do some quick shuffle to get past Rita. There are only three members still in Lafayette – Scott, Patrick, and I. (Had to make a call on whether the timeline should show Lafayette or Ontario. I am writing from Lafayette. But its about the team in Ontario, so I chose Ontario.)Incidentally, Continental Express did finally cancel the flights on Friday too, the eve of the hurricane. Had they been a little more flexible, rather thoughtful, about their policies we’d have still used their tickets. Per their policy, you could reroute without penalty or extra charge so long as you changed neither the origin nor destination. So if you were scheduled to fly out of Lafayette, you could not choose to drive to Baton Rouge and fly out a day early when Houston airport was still open. In order to do that you’d have to pay $700 to change each ticket. This policy stood even after flights from Lafayette were canceled. Like I mentioned before Continental would make even FEMA blush.

Thank goodness the State of Louisiana has some pre-negotiated rates, though out of a few limited city. I was able to get fully refundable tickets at very reasonable rates. It did require the team to drive some distance to Shreveport. Given the circumstances, that was the least of the concern. As for the non-refundable tickets, they became automatically refundable when Continental canceled the flights.

Having heard the stories of evacuees from Houston stranded on highway, I really feel lucky that the CajunBot Chase Team made it to Ontario. They did get caught up in the traffic of early evacuees, and took four hours to go around Houston. I thought four hours was long. But it doesn’t look so bad considering that later evacuees spent over 10, and sometimes 20, hours heading out of Houston.

And now Rita has come and gone. Areas West of Lafayette have really bore the brunt of the storm. Lafayette, at least my neighborhood, got a lot of water and high winds, but all the houses, roofs, and cars are all in tact. No fallen trees, except the crepe myrtle bushes outside my house. This was bad enough news for my wife, and I had to ‘do time’ shoveling dirt to anchor the bushes. No regrets at that. At least I did not have to climb the roof to put tarp like one of the last hurricane (I think Lily) hit.

With weather stabilizing and life coming back to order, its time to schedule flight to head west. DARPA has some ceremonies (meetings) that say ‘Attendance of Team Leader is mandatory’. Its not worth trying to find out what happens if Team Leader is stranded due to hurricane.

– Arun Lakhotia

Team changes itinerary to beat Hurricane Rita

Wednesday, September 21st, 2005

Lafayette, LA – If you are holding your breath whether Team CajunBot will make it past Hurricane Rita, let go. We have changed our itinerary to beat Rita.

We were scheduled to fly out Friday night from Lafayette, through Houston, to Ontario, CA.

Instead, Murali is en route as I write, about to reach Ontario in an hour or so, if all has gone well. Joshua, Pablo, Vidhya, Christopher, Santhosh, and Daro are driving to Shreveport, to catch the Friday early morning flight. The four hour drive in normal days, is likely to take over 10 hours, with everyone evacuating north.

The people still remaining in Lafayette are Patrick, Scott, and I; stayed back to take care of family as the storm looms. If all goes well, we will head west on Tuesday.

My many thanks to Shaleen “Shae” Landry of Navigant for finding ways to get the team out. She spent the first half of the day finding alternatives through Houston. Just when she had everything in place, we found Continental Express had shutdown today (Thursday’s flight).  She had to start all over again to get seats on American Airlines, flying out of Shreveport.

This experience brought out an interesting ideosyncracy of airlines that would make FEMA blush. Today, Thursday, around 11:00am Continental Express canceled its PM flights from Lafayette to Houston. I called to find out the status of the Friday flights, when we were scheduled to leave. Found that the Friday flight were expected to leave as scheduled. Imagine that. Flights a day before the storm were canceled, but the flights on the day (rather hours) of the storm were not.

Of course, according to their policy, I could not change to another carrier until they canceled the flights. So I am expected to wait and re-route my team after the storm hits and all flights, including those going north, are canceled.

Shae and I figured we will just buy new tickets on another airlines, instead of changing existing tickets. Once Continental decides it will cancel the ticket, we will seek refund for the tickets on canceled flights.

– Arun Lakhotia