All videos are in MPEG-4 format. They can be viewed with Quicktime, mplayer, VLC...

(3 min 30 sec sequence - comments in english)

This video presents the european Comets IST project final demonstration, that took place in Lousa (Portugal) in May 2005. The demo involves two helicopters and a blimp in a fire detection, confirmation and monitoring scenario. The video depicts the overall system architecture and the UAVs architecture.

(2 min 12 sec sequence - no sound)

This video shows Karma flying along with Heliv and Marvin over a fire, during the second yearly experiments of the european Comets project, in the airfield of Lousa near Coimbra, Portugal.

(3 min 47 sec sequence - no sound)

This video was recorded in May 2003, during some experiments of the european Comets project, organized by ADAI in the airfield of Lousa near Coimbra, Portugal. Karma has been totally upgraded (new enveloppe, new gondola, new motors), but still flies here under the control of a pilot. There was a wind of about 10 km/h blowing, which is the upper limit Karma can bear with: its trajectories are not very stable.

(1 min 13 sec sequence - no sound)

This movie shows the building of a digital elevation map during a tethered flight over the garden and parking of the lab. This model is built on the sole basis of the on-board stereovision bench images: no gyros, IMU or GPS data has been used here. The blimp is localized with a Simultaneous Localization And Mapping (SLAM) algorithm: landmarks are detected and localized as it moves, and simultaneously used to estimate the blimp position (the landmarks are shown with an ellipsoid that represents their uncertainty on the video).

Related publications

(1 min 35 sec sequence - no sound)

This video was recorded during the test flights of Karma in the airfield of Graulhet, near Toulouse, in August 2002. The blimp is remotely controlled by an operator, while the on-board CPU only gathers the data measured by the GPS, the two inclinometers, the compass and the two cameras of the stereo bench.
Karma flies along three loops in this sequence, at altitudes ranging from a few meters up to about 25 meters. At the end of the third loop, its speed is almost reduced to zero.

Related publications

(4 min sequence - comments in English)

We find this one pretty neat (it actually took us some time to achieve it :-). It shows Lama reaching a distant goal, adapting its navigation mode according to the kind of terrain it traverses. Two local navigation modes are depicted, and the way they are driven by a navigation planner that sets the sub-goal to reach is shown. While navigating, Lama localizes itself thanks to the visual motion estimation technique.

Related publications

(1 min 20 sec sequence - no sound)

This video illustrates the results of the visual motion estimation algorithm (shown in green) with the rover Lama during a 50 m long loop, compared with the position estimated by odometry (in red). Worth to notice is the behavior of both estimators when the rover slips laterally on the small rocky hill, and the estimated positions when the rover comes back to its starting point. The visual motion estimator undoubtedly beats odometry.

Related publications

(3 min 20 sec sequence - no sound)

This one is not very illustrative, as it only shows Lama roving without any visual explanation of the involved algorithms. It was recorded in the garden of the Cité de l'Espace museum of Toulouse, during a week-end organized within the temporary "Destination Mars" exposition. It was the first time that Lama evolved autonomously in public: very impressed, Lama was not in its best shape, and only accepted to evolve slowly, wavering along its paths.

Related publications

(3 min sequence - comments in english)

This film illustrates how Lama can avoid sparse obstacles, using a potential field based reactive navigation approach. The principle of the approach is depicted, and an exemple of a nice obstacle avoidance is shown.

Related publications

(1 min 30 sec sequence - comments in english)

This one is now an oldie! It is extracted from a nice TV report entitled "Cars on Mars", realized and broadcasted by the BBC in the "Tomorrow's world" programme.
It shows Adam reaching autonomously a visual target, modelling its environment with a home made 3D scanning laser range finder, and planning trajectories on the built digital elevation map. At that time, it took 3 minutes to gather a 100x100 points 3D image...

Related publications