Expert system based dynamic planner

During an inspection, several unforeseen situations might appear like the detection by the FLS of a fishing net, or a complex pattern shape of more than one pipeline over the seabed recognized by the MBE, or simply a detour due to obstacle detection. In these cases, it is hoped that the DMP module exhibits an "intelligent" behavior. To cope with these real situations in the marine world, it was resorted to the experience and skills of ROV operators. A little part of their knowledge was elicited and codified in the form of a real time expert

Fig. 4. Photographs of the low-cost ICTIOBOT prototype

system, the EN4AUV introduced earlier. It has been developed using CLIPS that constitutes a C language based shell, and allows the knowledge representation to be in the form of rules and frames (COOL or Clips Object Oriented Language). These formalisms are used in the knowledge base (KB) to represent the involved knowledge. The main feature of the expert system is to assess a current situation in order to act accordingly, in a clearly data driven/ reactive behavior. Thus, EN4AUV is a reactive expert system, taking the proper action for every different situation, and considering the pipeline/cable status, the type of survey, the different mission settings, and others.

These situations were coded as possible scenarios in about fifty rules, like the one presented in figure 5. As the knowledge about different situations increases, the knowledge base describing new scenarios can be completed an updated, yielding an incremental KB growth.

Each scenario triggers different searching or tracking strategies, which are then subtasks with their own features. Scenarios are based mainly in two ideas: a) the survey type, and b) tracking states. The first one is defined a priori in the SMP module, to establish the number of pipeline/cables to be tracked, the navigation depth, and other mission features. The other basic component of the scenario determination is the tracking state that changes when the SFM module updates its sensors. From this information, the EN4AUV is able to decide the status of the pipeline/cable, that is, if buried, exposed, intermittent or free-span, and how is the AUV as regard as the pipeline/cable, that is, if avoiding an obstacle, with the object under study considered as found or lost, or returning to a previous known position.

ins <- (iljjecl (is-a WORK IIWJCENARIW (Count Heacq?cr) (Search results?^ (Movie (?MOVfE) (Rjiicwj.latus'iFS))

?ev !■ [object (is-s SURVEY} [Mat Reactjure Witt]) (test ?tr ?m*)J (test (9q?FS LOST]] (test feq [send [OBJ STUDY] get Pfasenijayoutjsiatus) N0T_D£1KTEDJWT_BURjED)) tesl f= 0 ftffl

(assal fCuirattjaenaho BC6J) (assal fPLSnoirsady)} (feted M

(inserts S^0VIE1SC3) (send ?iVS(juJ-Movie t 'MOVIE) ttetactfty fa avoid the assert™ of multpte scenarios in tne same Kfi query

(printout ("CLIPSMAGHINE: F50 5 1 Current Scenario is SCS -searching a pipatrunkline" crit)

(prinflutt'GLIPSMACHNE inserting movie' S7M0VIE

Fig. 5. A rule from the knowledge-base of the DMP for the 8th scenario determination, in the typical CLIPS syntax

Once scenarios are established, a typical situation assessment task, EN4AUV must output a desired trajectory or must decide a pipeline/cable reacquisition.

To yield a desired trajectory, the actions are organized in a set o few simple subtasks: findstart, search, back to start, skip, and track. Then the final trajectory of the AUV is built by one of these subtasks, or by a concatenation of them. Examples of scenarios developed for the sea trials were the following ones:

• 1st Scenario: The AUV is tracking an exposed pipeline, navigating on top, at a fixed offset smaller or equal than 5 meters. Both the MBE and the MAG can detect it.

• 2nd Scenario: The AUV is tracking a buried pipeline on top, at a fixed offset, smaller or equal than 5 meters. The MBE may not be able to detect it, but the MAG can track it anyway.

• 3rd Scenario: The AUV is tracking an intermittently exposed and buried pipeline at a fixed offset. This is a sequence of alternative appearance of scenarios number one and two.

• 4th Scenario: The AUV is tracking a free-span pipeline at a fixed offset. The pipe is tracked mainly based on MBE readings, which may be detecting the pipe itself or the trench.

• 5th Scenario: The AUV is tracking a pipeline in the presence of one or more pipes (like infield pipelines) or other magnetic objects in the area. Measures from MBE as well as MAG are needed.

• 6th Scenario: The AUV is tracking a pipeline but avoiding an obstacle. In such scenario the certainty error may increase beyond its thresholds, but the EN4AUV knows where the pipe is and ignores the pipe_lost flag. The path planner module outputs a flag indicating this condition and the EN4AUV may query the legacy data to confirm the existence of an exclusion zone. Although sensor readings are not reliable, they are not turned off to be ready when the AUV is again over the pipeline.

• 7th Scenario: The AUV is searching a buried pipeline. No readings from MBE, just MAG will yield detection when the AUV is right over the pipe. With two detection (crossing) points the pipeline direction vector is computed and the AUV starts tracking from the last known point with this direction.

• 8th Scenario: The AUV is searching the pipeline, which is considered as lost. EN4AUV shall have an estimate of the trajectory from SFM considering the whole inputs: MBE, MAG, and LD. If not, it will estimate the initial parameters from mission settings and LD.

• 9th Scenario: The AUV is searching a pipeline in the presence of one or more pipes (like infield pipelines) or other magnetic objects in the area. Every information source is operative to discriminate the target under study (MBE, MAG, and LD).

• 10th Scenario: The AUV is skipping from one point to another. MBE, and MAG are off to save energy. This special situation appears when changing from one pipe to another to track, or from one zone of interest to another over the same pipeline.

• 11th Scenario: The AUV is going back to the last known position to start tracking, after founding the pipeline as a consequence of a successful search. MBE, and MAG are off.

• 12th Scenario: The maximum number of reacquisition after unsuccessful searches was reached. The mission is ended with a failure message.

• 13th Scenario: The AUV is tracking an exposed pipeline, navigating on top, at a fixed z_offset greater than 5 meters. The detection is done mainly with the MBE.

• 14th Scenario: The AUV is tracking a buried pipeline on top, at a fixed z_offset greater than 5 meters. The blind tracking is done mainly based on legacy data, and cannot last more than half a minute. After this, if there are no more sensor readings, a new search must be started.

Was this article helpful?

0 0
Digital Photography Mastery

Digital Photography Mastery

Insider Secrets Revealed By the Pro Showing You How to Become a Professional Photographer! Discover The Secret Tips & Techniques On How To Be A Professional Photographer, Start Producing High Quality Pictures and Skyrocket Your Photography Business Income Revenue To The Roof TODAY! You're About to Discover the Powerful Strategies and Method to Start Taking Sharp, Clear and High Quality Pictures Like the Professional Photographer Without Paying a Single Penny to the Expert!

Get My Free Ebook

Post a comment