1 | /*
|
---|
2 | * Licensed to the Apache Software Foundation (ASF) under one or more
|
---|
3 | * contributor license agreements. See the NOTICE file distributed with
|
---|
4 | * this work for additional information regarding copyright ownership.
|
---|
5 | * The ASF licenses this file to You under the Apache License, Version 2.0
|
---|
6 | * (the "License"); you may not use this file except in compliance with
|
---|
7 | * the License. You may obtain a copy of the License at
|
---|
8 | *
|
---|
9 | * http://www.apache.org/licenses/LICENSE-2.0
|
---|
10 | *
|
---|
11 | * Unless required by applicable law or agreed to in writing, software
|
---|
12 | * distributed under the License is distributed on an "AS IS" BASIS,
|
---|
13 | * WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
---|
14 | * See the License for the specific language governing permissions and
|
---|
15 | * limitations under the License.
|
---|
16 | */
|
---|
17 |
|
---|
18 | package agents.anac.y2019.harddealer.math3.ode.events;
|
---|
19 |
|
---|
20 | import agents.anac.y2019.harddealer.math3.RealFieldElement;
|
---|
21 | import agents.anac.y2019.harddealer.math3.ode.FieldODEState;
|
---|
22 | import agents.anac.y2019.harddealer.math3.ode.FieldODEStateAndDerivative;
|
---|
23 |
|
---|
24 | /** This interface represents a handler for discrete events triggered
|
---|
25 | * during ODE integration.
|
---|
26 | *
|
---|
27 | * <p>Some events can be triggered at discrete times as an ODE problem
|
---|
28 | * is solved. This occurs for example when the integration process
|
---|
29 | * should be stopped as some state is reached (G-stop facility) when the
|
---|
30 | * precise date is unknown a priori, or when the derivatives have
|
---|
31 | * discontinuities, or simply when the user wants to monitor some
|
---|
32 | * states boundaries crossings.
|
---|
33 | * </p>
|
---|
34 | *
|
---|
35 | * <p>These events are defined as occurring when a <code>g</code>
|
---|
36 | * switching function sign changes.</p>
|
---|
37 | *
|
---|
38 | * <p>Since events are only problem-dependent and are triggered by the
|
---|
39 | * independent <i>time</i> variable and the state vector, they can
|
---|
40 | * occur at virtually any time, unknown in advance. The integrators will
|
---|
41 | * take care to avoid sign changes inside the steps, they will reduce
|
---|
42 | * the step size when such an event is detected in order to put this
|
---|
43 | * event exactly at the end of the current step. This guarantees that
|
---|
44 | * step interpolation (which always has a one step scope) is relevant
|
---|
45 | * even in presence of discontinuities. This is independent from the
|
---|
46 | * stepsize control provided by integrators that monitor the local
|
---|
47 | * error (this event handling feature is available for all integrators,
|
---|
48 | * including fixed step ones).</p>
|
---|
49 | *
|
---|
50 | * @param <T> the type of the field elements
|
---|
51 | * @since 3.6
|
---|
52 | */
|
---|
53 | public interface FieldEventHandler<T extends RealFieldElement<T>> {
|
---|
54 |
|
---|
55 | /** Initialize event handler at the start of an ODE integration.
|
---|
56 | * <p>
|
---|
57 | * This method is called once at the start of the integration. It
|
---|
58 | * may be used by the event handler to initialize some internal data
|
---|
59 | * if needed.
|
---|
60 | * </p>
|
---|
61 | * @param initialState initial time, state vector and derivative
|
---|
62 | * @param finalTime target time for the integration
|
---|
63 | */
|
---|
64 | void init(FieldODEStateAndDerivative<T> initialState, T finalTime);
|
---|
65 |
|
---|
66 | /** Compute the value of the switching function.
|
---|
67 |
|
---|
68 | * <p>The discrete events are generated when the sign of this
|
---|
69 | * switching function changes. The integrator will take care to change
|
---|
70 | * the stepsize in such a way these events occur exactly at step boundaries.
|
---|
71 | * The switching function must be continuous in its roots neighborhood
|
---|
72 | * (but not necessarily smooth), as the integrator will need to find its
|
---|
73 | * roots to locate precisely the events.</p>
|
---|
74 | * <p>Also note that the integrator expect that once an event has occurred,
|
---|
75 | * the sign of the switching function at the start of the next step (i.e.
|
---|
76 | * just after the event) is the opposite of the sign just before the event.
|
---|
77 | * This consistency between the steps <string>must</strong> be preserved,
|
---|
78 | * otherwise {@link agents.anac.y2019.harddealer.math3.exception.NoBracketingException
|
---|
79 | * exceptions} related to root not being bracketed will occur.</p>
|
---|
80 | * <p>This need for consistency is sometimes tricky to achieve. A typical
|
---|
81 | * example is using an event to model a ball bouncing on the floor. The first
|
---|
82 | * idea to represent this would be to have {@code g(t) = h(t)} where h is the
|
---|
83 | * height above the floor at time {@code t}. When {@code g(t)} reaches 0, the
|
---|
84 | * ball is on the floor, so it should bounce and the typical way to do this is
|
---|
85 | * to reverse its vertical velocity. However, this would mean that before the
|
---|
86 | * event {@code g(t)} was decreasing from positive values to 0, and after the
|
---|
87 | * event {@code g(t)} would be increasing from 0 to positive values again.
|
---|
88 | * Consistency is broken here! The solution here is to have {@code g(t) = sign
|
---|
89 | * * h(t)}, where sign is a variable with initial value set to {@code +1}. Each
|
---|
90 | * time {@link #eventOccurred(FieldODEStateAndDerivative, boolean) eventOccurred}
|
---|
91 | * method is called, {@code sign} is reset to {@code -sign}. This allows the
|
---|
92 | * {@code g(t)} function to remain continuous (and even smooth) even across events,
|
---|
93 | * despite {@code h(t)} is not. Basically, the event is used to <em>fold</em>
|
---|
94 | * {@code h(t)} at bounce points, and {@code sign} is used to <em>unfold</em> it
|
---|
95 | * back, so the solvers sees a {@code g(t)} function which behaves smoothly even
|
---|
96 | * across events.</p>
|
---|
97 |
|
---|
98 | * @param state current value of the independent <i>time</i> variable, state vector
|
---|
99 | * and derivative
|
---|
100 | * @return value of the g switching function
|
---|
101 | */
|
---|
102 | T g(FieldODEStateAndDerivative<T> state);
|
---|
103 |
|
---|
104 | /** Handle an event and choose what to do next.
|
---|
105 |
|
---|
106 | * <p>This method is called when the integrator has accepted a step
|
---|
107 | * ending exactly on a sign change of the function, just <em>before</em>
|
---|
108 | * the step handler itself is called (see below for scheduling). It
|
---|
109 | * allows the user to update his internal data to acknowledge the fact
|
---|
110 | * the event has been handled (for example setting a flag in the {@link
|
---|
111 | * agents.anac.y2019.harddealer.math3.ode.FirstOrderDifferentialEquations
|
---|
112 | * differential equations} to switch the derivatives computation in
|
---|
113 | * case of discontinuity), or to direct the integrator to either stop
|
---|
114 | * or continue integration, possibly with a reset state or derivatives.</p>
|
---|
115 |
|
---|
116 | * <ul>
|
---|
117 | * <li>if {@link Action#STOP} is returned, the step handler will be called
|
---|
118 | * with the <code>isLast</code> flag of the {@link
|
---|
119 | * agents.anac.y2019.harddealer.math3.ode.sampling.StepHandler#handleStep handleStep}
|
---|
120 | * method set to true and the integration will be stopped,</li>
|
---|
121 | * <li>if {@link Action#RESET_STATE} is returned, the {@link #resetState
|
---|
122 | * resetState} method will be called once the step handler has
|
---|
123 | * finished its task, and the integrator will also recompute the
|
---|
124 | * derivatives,</li>
|
---|
125 | * <li>if {@link Action#RESET_DERIVATIVES} is returned, the integrator
|
---|
126 | * will recompute the derivatives,
|
---|
127 | * <li>if {@link Action#CONTINUE} is returned, no specific action will
|
---|
128 | * be taken (apart from having called this method) and integration
|
---|
129 | * will continue.</li>
|
---|
130 | * </ul>
|
---|
131 |
|
---|
132 | * <p>The scheduling between this method and the {@link
|
---|
133 | * agents.anac.y2019.harddealer.math3.ode.sampling.FieldStepHandler FieldStepHandler} method {@link
|
---|
134 | * agents.anac.y2019.harddealer.math3.ode.sampling.FieldStepHandler#handleStep(
|
---|
135 | * agents.anac.y2019.harddealer.math3.ode.sampling.FieldStepInterpolator, boolean)
|
---|
136 | * handleStep(interpolator, isLast)} is to call this method first and
|
---|
137 | * <code>handleStep</code> afterwards. This scheduling allows the integrator to
|
---|
138 | * pass <code>true</code> as the <code>isLast</code> parameter to the step
|
---|
139 | * handler to make it aware the step will be the last one if this method
|
---|
140 | * returns {@link Action#STOP}. As the interpolator may be used to navigate back
|
---|
141 | * throughout the last step, user code called by this method and user
|
---|
142 | * code called by step handlers may experience apparently out of order values
|
---|
143 | * of the independent time variable. As an example, if the same user object
|
---|
144 | * implements both this {@link FieldEventHandler FieldEventHandler} interface and the
|
---|
145 | * {@link agents.anac.y2019.harddealer.math3.ode.sampling.FieldStepHandler FieldStepHandler}
|
---|
146 | * interface, a <em>forward</em> integration may call its
|
---|
147 | * {code eventOccurred} method with t = 10 first and call its
|
---|
148 | * {code handleStep} method with t = 9 afterwards. Such out of order
|
---|
149 | * calls are limited to the size of the integration step for {@link
|
---|
150 | * agents.anac.y2019.harddealer.math3.ode.sampling.FieldStepHandler variable step handlers}.</p>
|
---|
151 |
|
---|
152 | * @param state current value of the independent <i>time</i> variable, state vector
|
---|
153 | * and derivative
|
---|
154 | * @param increasing if true, the value of the switching function increases
|
---|
155 | * when times increases around event (note that increase is measured with respect
|
---|
156 | * to physical time, not with respect to integration which may go backward in time)
|
---|
157 | * @return indication of what the integrator should do next, this
|
---|
158 | * value must be one of {@link Action#STOP}, {@link Action#RESET_STATE},
|
---|
159 | * {@link Action#RESET_DERIVATIVES} or {@link Action#CONTINUE}
|
---|
160 | */
|
---|
161 | Action eventOccurred(FieldODEStateAndDerivative<T> state, boolean increasing);
|
---|
162 |
|
---|
163 | /** Reset the state prior to continue the integration.
|
---|
164 |
|
---|
165 | * <p>This method is called after the step handler has returned and
|
---|
166 | * before the next step is started, but only when {@link
|
---|
167 | * #eventOccurred(FieldODEStateAndDerivative, boolean) eventOccurred} has itself
|
---|
168 | * returned the {@link Action#RESET_STATE} indicator. It allows the user to reset
|
---|
169 | * the state vector for the next step, without perturbing the step handler of the
|
---|
170 | * finishing step. If the {@link #eventOccurred(FieldODEStateAndDerivative, boolean)
|
---|
171 | * eventOccurred} never returns the {@link Action#RESET_STATE} indicator, this
|
---|
172 | * function will never be called, and it is safe to leave its body empty.</p>
|
---|
173 | * @param state current value of the independent <i>time</i> variable, state vector
|
---|
174 | * and derivative
|
---|
175 | * @return reset state (note that it does not include the derivatives, they will
|
---|
176 | * be added automatically by the integrator afterwards)
|
---|
177 | */
|
---|
178 | FieldODEState<T> resetState(FieldODEStateAndDerivative<T> state);
|
---|
179 |
|
---|
180 | }
|
---|