Sample TWCC implementation (#1957)

* encoder bitrate change based on twcc

* Change to 5% inc and dec

* modify percentages

* ema based calc

* EMa fix

* Nits

* Readme

* flip

* memset remove

* Readme update, move enable flags to createSampleConfiguration

* Add codecov token

Author: disa6302

Diff for: .github/workflows/codecov.yml

@@ -13,6 +13,7 @@ jobs:
     runs-on: ubuntu-20.04
     env:
       AWS_KVS_LOG_LEVEL: 2
+      CODECOV_TOKEN: ${{ secrets.CODECOV_TOKEN }}
     permissions:
       id-token: write
       contents: read

Diff for: README.md

@@ -396,21 +396,32 @@ createLwsIotCredentialProvider(
 freeIotCredentialProvider(&pSampleConfiguration->pCredentialProvider);
 ```
 
-## Use of TWCC
-In order to listen in on TWCC reports, the application must set up a callback using the `peerConnectionOnSenderBandwidthEstimation` API. In our samples, it is set up like this:
+## TWCC support
+
+Transport Wide Congestion Control (TWCC) is a mechanism in WebRTC designed to enhance the performance and reliability of real-time communication over the internet. TWCC addresses the challenges of network congestion by providing detailed feedback on the transport of packets across the network, enabling adaptive bitrate control and optimization of media streams in real-time. This feedback mechanism is crucial for maintaining high-quality audio and video communication, as it allows senders to adjust their transmission strategies based on comprehensive information about packet losses, delays, and jitter experienced across the entire transport path.
+
+The importance of TWCC in WebRTC lies in its ability to ensure efficient use of available network bandwidth while minimizing the negative impacts of network congestion. By monitoring the delivery of packets across the network, TWCC helps identify bottlenecks and adjust the media transmission rates accordingly. This dynamic approach to congestion control is essential for preventing degradation in call quality, such as pixelation, stuttering, or drops in audio and video streams, especially in environments with fluctuating network conditions.
+
+To learn more about TWCC, check [TWCC spec](https://datatracker.ietf.org/doc/html/draft-holmer-rmcat-transport-wide-cc-extensions-01)
+
+### Enabling TWCC support
+
+TWCC is enabled by default in the SDK samples (via `pSampleConfiguration->enableTwcc`) flag. In order to disable it, set this flag to `FALSE`.
 
 ```c
-CHK_STATUS(peerConnectionOnSenderBandwidthEstimation(pSampleStreamingSession->pPeerConnection, (UINT64) pSampleStreamingSession,
-                                                     sampleSenderBandwidthEstimationHandler));
+pSampleConfiguration->enableTwcc = FALSE;
 ```
 
-Note that TWCC is disabled by default in the SDK samples. In order to enable it, set the `disableSenderSideBandwidthEstimation` flag to FALSE. For example,
-
+If not using the samples directly, 2 things need to be done to set up Twcc:
+1. Set the `disableSenderSideBandwidthEstimation` to `FALSE`:
 ```c
-RtcConfiguration configuration;
 configuration.kvsRtcConfiguration.disableSenderSideBandwidthEstimation = FALSE;
 ```
-
+2. Set the callback that will have the business logic to modify the bitrate based on packet loss information. The callback can be set using `peerConnectionOnSenderBandwidthEstimation()`:
+```c
+CHK_STATUS(peerConnectionOnSenderBandwidthEstimation(pSampleStreamingSession->pPeerConnection, (UINT64) pSampleStreamingSession,
+                                                     sampleSenderBandwidthEstimationHandler));
+```
 
 ## Use Pre-generated Certificates
 The certificate generating function ([createCertificateAndKey](https://awslabs.github.io/amazon-kinesis-video-streams-webrtc-sdk-c/Dtls__openssl_8c.html#a451c48525b0c0a8919a880d6834c1f7f)) in createDtlsSession() can take between 5 - 15 seconds in low performance embedded devices, it is called for every peer connection creation when KVS WebRTC receives an offer. To avoid this extra start-up latency, certificate can be pre-generated and passed in when offer comes.
@@ -510,6 +521,32 @@ To disable threadpool, run `cmake .. -DENABLE_KVS_THREADPOOL=OFF`
 ### Thread stack sizes
 The default thread stack size for the KVS WebRTC SDK is 64 kb. Notable stack sizes that may need to be changed for your specific application will be the ConnectionListener Receiver thread and the media sender threads. Please modify the stack sizes for these media dependent threads to be suitable for the media your application is processing.
 
+### Set up TWCC
+TWCC is a mechanism in WebRTC designed to enhance the performance and reliability of real-time communication over the Internet. TWCC addresses the challenges of network congestion by providing detailed feedback on the transport of packets across the network, enabling adaptive bitrate control and optimization of 
+media streams in real-time. This feedback mechanism is crucial for maintaining high-quality audio and video communication, as it allows senders to adjust their transmission strategies based on comprehensive information about packet losses, delays, and jitter experienced across the entire transport path.
+The importance of TWCC in WebRTC lies in its ability to ensure efficient use of available network bandwidth while minimizing the negative impacts of network congestion. By monitoring the delivery of packets across the network, TWCC helps identify bottlenecks and adjust the media transmission rates accordingly. 
+This dynamic approach to congestion control is essential for preventing degradation in call quality, such as pixelation, stuttering, or drops in audio and video streams, especially in environments with fluctuating network conditions. To learn more about TWCC, you can refer to the [RFC draft](https://datatracker.ietf.org/doc/html/draft-holmer-rmcat-transport-wide-cc-extensions-01)
+
+In order to enable TWCC usage in the SDK, 2 things need to be set up:
+
+1. Set the `disableSenderSideBandwidthEstimation` to FALSE. In our samples, the value is set using `enableTwcc` flag in `pSampleConfiguration`
+
+```c
+pSampleConfiguration->enableTwcc = TRUE; // to enable TWCC
+pSampleConfiguration->enableTwcc = FALSE; // to disable TWCC
+configuration.kvsRtcConfiguration.disableSenderSideBandwidthEstimation = !pSampleConfiguration->enableTwcc;
+```
+
+2. Set the callback that will have the business logic to modify the bitrate based on packet loss information. The callback can be set using `peerConnectionOnSenderBandwidthEstimation()`.
+
+```c
+CHK_STATUS(peerConnectionOnSenderBandwidthEstimation(pSampleStreamingSession->pPeerConnection, (UINT64) pSampleStreamingSession,
+                                                     sampleSenderBandwidthEstimationHandler));
+```
+
+By default, our SDK enables TWCC listener. The SDK has a sample implementation to integrate TWCC into the Gstreamer pipeline via the `sampleSenderBandwidthEstimationHandler` callback. To get more details, look for this specific callback.
+
+
 ### Setting ICE related timeouts
 
 There are some default timeout values set for different steps in ICE in the [KvsRtcConfiguration](https://awslabs.github.io/amazon-kinesis-video-streams-webrtc-sdk-c/structKvsRtcConfiguration.html). These are configurable in the application. While the defaults are generous, there could be applications that might need more flexibility to improve chances of connection establishment because of poor network. 

Diff for: samples/Common.c

@@ -414,13 +414,13 @@ STATUS initializePeerConnection(PSampleConfiguration pSampleConfiguration, PRtcP
     configuration.kvsRtcConfiguration.iceSetInterfaceFilterFunc = NULL;
 
     // disable TWCC
-    configuration.kvsRtcConfiguration.disableSenderSideBandwidthEstimation = TRUE;
+    configuration.kvsRtcConfiguration.disableSenderSideBandwidthEstimation = !(pSampleConfiguration->enableTwcc);
+    DLOGI("TWCC is : %s", configuration.kvsRtcConfiguration.disableSenderSideBandwidthEstimation ? "Disabled" : "Enabled");
 
     // Set the ICE mode explicitly
     configuration.iceTransportPolicy = ICE_TRANSPORT_POLICY_ALL;
 
     configuration.kvsRtcConfiguration.enableIceStats = pSampleConfiguration->enableIceStats;
-    configuration.kvsRtcConfiguration.disableSenderSideBandwidthEstimation = TRUE;
     // Set the  STUN server
     PCHAR pKinesisVideoStunUrlPostFix = KINESIS_VIDEO_STUN_URL_POSTFIX;
     // If region is in CN, add CN region uri postfix
@@ -554,10 +554,12 @@ STATUS createSampleStreamingSession(PSampleConfiguration pSampleConfiguration, P
 
     ATOMIC_STORE_BOOL(&pSampleStreamingSession->terminateFlag, FALSE);
     ATOMIC_STORE_BOOL(&pSampleStreamingSession->candidateGatheringDone, FALSE);
-
     pSampleStreamingSession->peerConnectionMetrics.peerConnectionStats.peerConnectionStartTime = GETTIME() / HUNDREDS_OF_NANOS_IN_A_MILLISECOND;
-    // Flag to enable SDK to calculate selected ice server, local, remote and candidate pair stats.
-    pSampleConfiguration->enableIceStats = FALSE;
+
+    if (pSampleConfiguration->enableTwcc) {
+        pSampleStreamingSession->twccMetadata.updateLock = MUTEX_CREATE(TRUE);
+    }
+
     CHK_STATUS(initializePeerConnection(pSampleConfiguration, &pSampleStreamingSession->pPeerConnection));
     CHK_STATUS(peerConnectionOnIceCandidate(pSampleStreamingSession->pPeerConnection, (UINT64) pSampleStreamingSession, onIceCandidateHandler));
     CHK_STATUS(
@@ -604,8 +606,10 @@ STATUS createSampleStreamingSession(PSampleConfiguration pSampleConfiguration, P
     CHK_STATUS(transceiverOnBandwidthEstimation(pSampleStreamingSession->pAudioRtcRtpTransceiver, (UINT64) pSampleStreamingSession,
                                                 sampleBandwidthEstimationHandler));
     // twcc bandwidth estimation
-    CHK_STATUS(peerConnectionOnSenderBandwidthEstimation(pSampleStreamingSession->pPeerConnection, (UINT64) pSampleStreamingSession,
-                                                         sampleSenderBandwidthEstimationHandler));
+    if (pSampleConfiguration->enableTwcc) {
+        CHK_STATUS(peerConnectionOnSenderBandwidthEstimation(pSampleStreamingSession->pPeerConnection, (UINT64) pSampleStreamingSession,
+                                                             sampleSenderBandwidthEstimationHandler));
+    }
     pSampleStreamingSession->startUpLatency = 0;
 CleanUp:
 
@@ -656,6 +660,12 @@ STATUS freeSampleStreamingSession(PSampleStreamingSession* ppSampleStreamingSess
     }
     MUTEX_UNLOCK(pSampleConfiguration->sampleConfigurationObjLock);
 
+    if (pSampleConfiguration->enableTwcc) {
+        if (IS_VALID_MUTEX_VALUE(pSampleStreamingSession->twccMetadata.updateLock)) {
+            MUTEX_FREE(pSampleStreamingSession->twccMetadata.updateLock);
+        }
+    }
+
     CHK_LOG_ERR(closePeerConnection(pSampleStreamingSession->pPeerConnection));
     CHK_LOG_ERR(freePeerConnection(&pSampleStreamingSession->pPeerConnection));
     SAFE_MEMFREE(pSampleStreamingSession);
@@ -706,27 +716,61 @@ VOID sampleBandwidthEstimationHandler(UINT64 customData, DOUBLE maximumBitrate)
     DLOGV("received bitrate suggestion: %f", maximumBitrate);
 }
 
+// Sample callback for TWCC. Average packet is calculated with exponential moving average (EMA). If average packet lost is <= 5%,
+// the current bitrate is increased by 5%. If more than 5%, the current bitrate
+// is reduced by percent lost. Bitrate update is allowed every second and is increased/decreased upto the limits
 VOID sampleSenderBandwidthEstimationHandler(UINT64 customData, UINT32 txBytes, UINT32 rxBytes, UINT32 txPacketsCnt, UINT32 rxPacketsCnt,
                                             UINT64 duration)
 {
-    UNUSED_PARAM(customData);
     UNUSED_PARAM(duration);
-    UNUSED_PARAM(rxBytes);
-    UNUSED_PARAM(txBytes);
+    UINT64 videoBitrate, audioBitrate;
+    UINT64 currentTimeMs, timeDiff;
     UINT32 lostPacketsCnt = txPacketsCnt - rxPacketsCnt;
-    UINT32 percentLost = lostPacketsCnt * 100 / txPacketsCnt;
-    UINT32 bitrate = 1024;
-    if (percentLost < 2) {
-        // increase encoder bitrate by 2 percent
-        bitrate *= 1.02f;
-    } else if (percentLost > 5) {
-        // decrease encoder bitrate by packet loss percent
-        bitrate *= (1.0f - percentLost / 100.0f);
-    }
-    // otherwise keep bitrate the same
-
-    DLOGV("received sender bitrate estimation: suggested bitrate %u sent: %u bytes %u packets received: %u bytes %u packets in %lu msec", bitrate,
-          txBytes, txPacketsCnt, rxBytes, rxPacketsCnt, duration / 10000ULL);
+    DOUBLE percentLost = (DOUBLE) ((txPacketsCnt > 0) ? (lostPacketsCnt * 100 / txPacketsCnt) : 0.0);
+    SampleStreamingSession* pSampleStreamingSession = (SampleStreamingSession*) customData;
+
+    if (pSampleStreamingSession == NULL) {
+        DLOGW("Invalid streaming session (NULL object)");
+        return;
+    }
+
+    // Calculate packet loss
+    pSampleStreamingSession->twccMetadata.averagePacketLoss =
+        EMA_ACCUMULATOR_GET_NEXT(pSampleStreamingSession->twccMetadata.averagePacketLoss, ((DOUBLE) percentLost));
+
+    currentTimeMs = GETTIME();
+    timeDiff = currentTimeMs - pSampleStreamingSession->twccMetadata.lastAdjustmentTimeMs;
+    if (timeDiff < TWCC_BITRATE_ADJUSTMENT_INTERVAL_MS) {
+        // Too soon for another adjustment
+        return;
+    }
+
+    MUTEX_LOCK(pSampleStreamingSession->twccMetadata.updateLock);
+    videoBitrate = pSampleStreamingSession->twccMetadata.currentVideoBitrate;
+    audioBitrate = pSampleStreamingSession->twccMetadata.currentAudioBitrate;
+
+    if (pSampleStreamingSession->twccMetadata.averagePacketLoss <= 5) {
+        // increase encoder bitrate by 5 percent with a cap at MAX_BITRATE
+        videoBitrate = (UINT64) MIN(videoBitrate * 1.05, MAX_VIDEO_BITRATE_KBPS);
+        // increase encoder bitrate by 5 percent with a cap at MAX_BITRATE
+        audioBitrate = (UINT64) MIN(audioBitrate * 1.05, MAX_AUDIO_BITRATE_BPS);
+    } else {
+        // decrease encoder bitrate by average packet loss percent, with a cap at MIN_BITRATE
+        videoBitrate = (UINT64) MAX(videoBitrate * (1.0 - pSampleStreamingSession->twccMetadata.averagePacketLoss / 100.0), MIN_VIDEO_BITRATE_KBPS);
+        // decrease encoder bitrate by average packet loss percent, with a cap at MIN_BITRATE
+        audioBitrate = (UINT64) MAX(audioBitrate * (1.0 - pSampleStreamingSession->twccMetadata.averagePacketLoss / 100.0), MIN_AUDIO_BITRATE_BPS);
+    }
+
+    // Update the session with the new bitrate and adjustment time
+    pSampleStreamingSession->twccMetadata.newVideoBitrate = videoBitrate;
+    pSampleStreamingSession->twccMetadata.newAudioBitrate = audioBitrate;
+    MUTEX_UNLOCK(pSampleStreamingSession->twccMetadata.updateLock);
+
+    pSampleStreamingSession->twccMetadata.lastAdjustmentTimeMs = currentTimeMs;
+
+    DLOGI("Adjustment made: average packet loss = %.2f%%, timediff: %llu ms", pSampleStreamingSession->twccMetadata.averagePacketLoss, timeDiff);
+    DLOGI("Suggested video bitrate %u kbps, suggested audio bitrate: %u bps, sent: %u bytes %u packets received: %u bytes %u packets in %lu msec",
+          videoBitrate, audioBitrate, txBytes, txPacketsCnt, rxBytes, rxPacketsCnt, duration / 10000ULL);
 }
 
 STATUS handleRemoteCandidate(PSampleStreamingSession pSampleStreamingSession, PSignalingMessage pSignalingMessage)
@@ -915,6 +959,12 @@ STATUS createSampleConfiguration(PCHAR channelName, SIGNALING_CHANNEL_ROLE_TYPE
     pSampleConfiguration->pregenerateCertTimerId = MAX_UINT32;
     pSampleConfiguration->signalingClientMetrics.version = SIGNALING_CLIENT_METRICS_CURRENT_VERSION;
 
+    // Flag to enable SDK to calculate selected ice server, local, remote and candidate pair stats.
+    pSampleConfiguration->enableIceStats = FALSE;
+
+    // Flag to enable/disable TWCC
+    pSampleConfiguration->enableTwcc = TRUE;
+
     ATOMIC_STORE_BOOL(&pSampleConfiguration->interrupted, FALSE);
     ATOMIC_STORE_BOOL(&pSampleConfiguration->mediaThreadStarted, FALSE);
     ATOMIC_STORE_BOOL(&pSampleConfiguration->appTerminateFlag, FALSE);

Diff for: samples/Samples.h

@@ -83,6 +83,12 @@ extern "C" {
 #define MAX_SIGNALING_CLIENT_METRICS_MESSAGE_SIZE 736 // strlen(SIGNALING_CLIENT_METRICS_JSON_TEMPLATE) + 20 * 10
 #define MAX_ICE_AGENT_METRICS_MESSAGE_SIZE        113 // strlen(ICE_AGENT_METRICS_JSON_TEMPLATE) + 20 * 2
 
+#define TWCC_BITRATE_ADJUSTMENT_INTERVAL_MS 1000 * HUNDREDS_OF_NANOS_IN_A_MILLISECOND
+#define MIN_VIDEO_BITRATE_KBPS              512     // Unit kilobits/sec. Value could change based on codec.
+#define MAX_VIDEO_BITRATE_KBPS              2048000 // Unit kilobits/sec. Value could change based on codec.
+#define MIN_AUDIO_BITRATE_BPS               4000    // Unit bits/sec. Value could change based on codec.
+#define MAX_AUDIO_BITRATE_BPS               650000  // Unit bits/sec. Value could change based on codec.
+
 typedef enum {
     SAMPLE_STREAMING_VIDEO_ONLY,
     SAMPLE_STREAMING_AUDIO_VIDEO,
@@ -160,6 +166,7 @@ typedef struct {
     PCHAR rtspUri;
     UINT32 logLevel;
     BOOL enableIceStats;
+    BOOL enableTwcc;
 } SampleConfiguration, *PSampleConfiguration;
 
 typedef struct {
@@ -179,6 +186,16 @@ typedef struct {
 
 typedef VOID (*StreamSessionShutdownCallback)(UINT64, PSampleStreamingSession);
 
+typedef struct {
+    MUTEX updateLock;
+    UINT64 lastAdjustmentTimeMs;
+    UINT64 currentVideoBitrate;
+    UINT64 currentAudioBitrate;
+    UINT64 newVideoBitrate;
+    UINT64 newAudioBitrate;
+    DOUBLE averagePacketLoss;
+} TwccMetadata, *PTwccMetadata;
+
 struct __SampleStreamingSession {
     volatile ATOMIC_BOOL terminateFlag;
     volatile ATOMIC_BOOL candidateGatheringDone;
@@ -198,6 +215,7 @@ struct __SampleStreamingSession {
     UINT64 startUpLatency;
     RtcMetricsHistory rtcMetricsHistory;
     BOOL remoteCanTrickleIce;
+    TwccMetadata twccMetadata;
 
     // this is called when the SampleStreamingSession is being freed
     StreamSessionShutdownCallback shutdownCallback;