explanation passed back by the scheduler, corresponds to one of the 9 types of variables listed in Table 3.
• Translation is further complicated by the fact that we have multiple resource allocation policies to help the planner in communicating the resource allocation preferences of the user to the scheduler. Different scheduling policies lead to different types of CSP encodings of the scheduling problem. As a result, we need different rules in translating the scheduler’s failure explanations back to the planner. The complexity of the translation process increases along with the complexity of the different classes. Regarding the second issue, for the simplest class, INFRES, each failure explanation (conflict set) resulting from solving the scheduling CSP can be directly converted to one set of action nodes in the planning graph of the planner. For class FIX, besides the variables corresponding to actions in the plan, the scheduling CSP also has variables representing the free/unfree actions. Therefore, we need to reason about such variables related to the spans, before we can convert the scheduler’s conflict set back to the set of actions in the planner. For the class SAMELEN, because the ranges of the possible position values for the actions in the scheduling CSP are widened, one variable in the scheduling CSP corresponds to many action nodes in the plan graph. Therefore, the translation rules for this class should be able to convert one conflict set of the scheduling CSP to many action sets in the planner CSP. Finally, for the INCRELEN class, because the ranges of the position of actions in the scheduler encoding (SE) are allowed to go beyond the highest level of the plan graph, there is no clear way to map them back to action sets in the planners. Therefore, we currently do not support the RealPlan-PP interactions between the two modules in this class.
To aid the translation, we start by keeping track of how the scheduler’s variables are derived from the resource spans in the causal plan. Suppose there is a span S i,j that is started and ended by the actions A i and A j . We know (see Section 4.1) that this single span gives raise to the variables, RA i , RA j , PA i and PA j . To convert the variables in the scheduler’s CSP to action values of the planner’s CSP, we thus invert this mapping as
follows:
อธิบายผ่านกลับ โดยตัวจัดกำหนดการ สอดคล้องกับชนิดของตัวแปรที่ระบุไว้ในตารางที่ 3 9• Translation is further complicated by the fact that we have multiple resource allocation policies to help the planner in communicating the resource allocation preferences of the user to the scheduler. Different scheduling policies lead to different types of CSP encodings of the scheduling problem. As a result, we need different rules in translating the scheduler’s failure explanations back to the planner. The complexity of the translation process increases along with the complexity of the different classes. Regarding the second issue, for the simplest class, INFRES, each failure explanation (conflict set) resulting from solving the scheduling CSP can be directly converted to one set of action nodes in the planning graph of the planner. For class FIX, besides the variables corresponding to actions in the plan, the scheduling CSP also has variables representing the free/unfree actions. Therefore, we need to reason about such variables related to the spans, before we can convert the scheduler’s conflict set back to the set of actions in the planner. For the class SAMELEN, because the ranges of the possible position values for the actions in the scheduling CSP are widened, one variable in the scheduling CSP corresponds to many action nodes in the plan graph. Therefore, the translation rules for this class should be able to convert one conflict set of the scheduling CSP to many action sets in the planner CSP. Finally, for the INCRELEN class, because the ranges of the position of actions in the scheduler encoding (SE) are allowed to go beyond the highest level of the plan graph, there is no clear way to map them back to action sets in the planners. Therefore, we currently do not support the RealPlan-PP interactions between the two modules in this class.เพื่อช่วยในการแปล เราเริ่มต้น โดยติดตามวิธีมาจากครอบคลุมทรัพยากรตัวจัดกำหนดการแปรแผนเชิงสาเหตุ สมมติว่ามีช่วง S i, j ที่จะเริ่มต้น และสิ้นสุดลง โดยการดำเนินการที่ผม และเจ เรารู้ (ดูหัวข้อ 4.1) ว่าเดี่ยวนี้ขยาย ให้เพิ่มกับตัวแปร RA i, RA j, PA ผมและ PA j การแปลงตัวแปรใน CSP ที่ตัวจัดกำหนดการการดำเนินการค่าของ CSP ของวางแผน เราจึงสลับการแมปนี้เป็นต่อไปนี้:
การแปล กรุณารอสักครู่..
