Architecture: sensors, actuators and intelligent behaviors

Homepage | Research | Students | Publications | Applications

I'm interested in generation of robot behaviors. This is a wide area with many challenging issues involved: environment estimation through sensor data, robust control, action selection, attention, etc. I'm focused in the perception and control architecture of the robot, which allows an autonomous robot to exhibit "intelligent" behaviors, reacting to different stimuli and accomplishing some commanded goals in a dynamic indoor environment. The architecture is a main issue, as it organizes all the robot sensor and actuation capabilities, in order to unfold them properly to generate behavior.

There are several robots I've worked with: B21 from RWI, RCX from LEGO, EyeBot, Pioneer, Xavier robot, etc. Even I've been exposed to Aibo from Sony. Always the same stuff: a set of sensors, a set of actuators and computing power in between. In the software side, an operating system and luckily enough, a programming environment to develop my own robotic applicactions.

jdeneoc: a component based C implementation of JDE
jde+: an object oriented implementation of JDE
jdec: a C implementation of JDE
JDE: dynamic schema hierarchy for behavior generation

jde+: an object oriented implementation of JDE

The JDE architecture has been also implemented in C++, gathering all the experience accumulated while using jde.c. A new design was carried out and implemented in a new software environment, jde+, which is closer to theorical proposal JDE than jde.c. It offers several advantages:

several instances of the same schema are allowed, even concurrent ones, each with its own modulation.
the schema interface is cleaner than jde.c. For instance, schema communication among brothers is strictly prohibited, only father-child communication is allowed and enforced.
most of the complexity of new schema programming is hided. The behavior designer only has to focus on schema-specific code (schema iteration, arbitration and preconditions). The general parts of its schemas are implemented by jde+ and inherited.
dynamic (an efficient) load of schemas is allowed at runtime, making true the reconfiguration capability promised by JDE.

jdec: a C implementation of JDE

jdec is a software development suite for robotics applications, written in C language. It provides a component oriented programming environment where the robot control program is made up of a collection of concurrent asynchronous threads named schemas. It is based on the JDE cognitive architecture for autonomous robots, described in my PhD thesis.

Inside jdec getting sensor measurements is as simple as reading a local variable, and ordering motor commands as easy as writing an actuator variable. The platform provides a shared variable API for the robot programming. It updates the sensor variables with fresh readings and periodically sends the commands in the actuator variables to the real motors. The robotic application reads and writes such variables to unfold its behavior. Each variable is tied to a given device (real, simulated or remote) in the configuration file. Current release provides drivers for sonar, SICK laser, encoders, firewire cameras, USB cameras, base motors, video files and pantilt unit. Drivers and schemas are inserted in the system as plugins. Jdec also includes several schemas ready to reuse: person following, visual attention, local navigation, global navigation, global localization, vision based wandering, laser segmentation, building occupancy grids, image filtering, etc.

Though jdec was born in 1997 as the software platform for a PhD dissertation, it is no longer a single person effort. Besides the user pool, a selected developer group actively maintains it and extends its functionality. Check the Official JDEC webpage for the latest release, contributions and more information.

JDE: dynamic schema hierarchy for behavior generation

I've developed an architecture that exploits the perceptual information and controls the robot movements. The architecture is named JDE and is organized as a hierarchy of reactive behaviors that are sensitive to goal modulation too (schemas). It takes inspiration from ethology hierarchies (Lorenz and Tinbergen studies of animal behavior). The system reacts fast to external estimuli due to low level control loops. High level agents can select the estimuli to which the low level agents should react, using the modulation parameters and biasing this way the behavior to goal-oriented movements.

architecture for person following

More details of it and its evolution in: